Subjec i e and Objec i e Quali y Assessmen o Dynamic
Poin Cloud wi h Visual A en ion in 6 DoF
XUEMEI ZHOU,Cen um Wiskunde & In o ma ica, Ams e dam, TU Del , Del , The Ne he lands
IRENE VIOLA,Cen um Wiskunde & In o ma ica, Ams e dam, The Ne he lands
EVANGELOS ALEXIOU,Xiaomi Technology, The Hague, The Ne he lands
JACK JANSEN,Cen um Wiskunde & In o ma ica, Ams e dam, The Ne he lands
PABLO CESAR,Cen um Wiskunde & In o ma ica, Ams e dam, TU Del , Del , The Ne he lands
Pe cep ual quali y assessmen o Dynamic Poin Cloud (DPC) con en s plays an impo an ole in a ious
Vi ual Reali y (VR) applica ions ha in ol e human beings as he end use . Unde s anding and modeling
pe cep ual quali y assessmen is g ea ly en iched by insigh s om isual a en ion. Howe e , inco po a ing
aspec s o isual a en ion in DPC quali y models is la gely unexplo ed, as g ound- u h isual a en ion da a
a e sca cely a ailable. Besides, es ing me hods and p ocedu es o collec ing isual a en ion da a a e s ill o
be ag eed on. This a icle p esen s a da ase con aining subjec i e opinion sco es and isual a en ion maps o
DPCs, collec ed in a VR en i onmen using eye- acking echnology. Bo h he quali y sco e and eye- acking
da a we e collec ed du ing a subjec i e quali y assessmen expe imen , in which subjec s we e ins uc ed o
wa ch and a e DPCs a a ious deg ada ion le els unde 6 Deg ees o F eedom (DoF) inspec ion, using a
head-moun ed display. Quali a i e in e iew analysis was also conduc ed a e he expe imen . The da ase
consis s o 50 DPCs, including 5 e e ence DPCs, wi h each e e ence encoded a 3 dis o ion le els using 3
di e en codecs (namely G-PCC, V-PCC, CWI-PCL), amoun ing o a o al o 9 deg aded e sion pe e e ence.
Addi ionally, i inco po a es 1,000 gaze ials om 40 pa icipan s, yielding a o al o 15,000 isual a en ion
maps ac oss all he DPCs.
We addi ionally benchma k objec i e quali y me ics o iginally designed o s a ic poin clouds, e alua ing
hei pe o mance in ou da ase using wo empo al pooling s a egies. Fu he mo e, we employ he isual
a en ion da a ha a e e ie ed du ing ou expe imen o e alua e whe he he pe o mance o widely used
objec i e quali y me ics is imp o ed by conside ing subjec i e measu emen s o isual a en ion. This da ase
es ablishes a link be ween quali y assessmen and isual a en ion wi hin he con ex o DPC. Mo eo e ,
hema ic analysis o he in e iews helps unco e use beha io and ac o s impac ing pe cep ual quali y o
DPC in 6 DoF. This wo k deepens ou unde s anding o DPC quali y assessmen and isual a en ion, d i ing
p og ess in he ealm o VR expe iences and pe cep ion.
CCS Concep s: • Human-cen e ed compu ing
→
Visualiza ion design and e alua ion me hods; •
Compu ing me hodologies
→
Pe cep ion;Model de elopmen and analysis;Image p ocessing;Vi ual
eali y;In e es poin and salien egion de ec ions;
This wo k was suppo ed h ough he NWO WISE g an and he Eu opean Commission Ho izon Eu ope p og am, unde
he g an ag eemen 101070109, TRANSMIXR h ps:// ansmix .eu/. Funded by he Eu opean Union.
Au ho s’ Con ac In o ma ion: Xuemei Zhou (co esponding au ho ),Cen um Wiskunde & In o ma ica, Ams e dam, TU
Del , Del , The Ne he lands; e-mail: [email p o ec ed]; I ene Viola,Cen um Wiskunde & In o ma ica, Ams e dam,
The Ne he lands; e-mail: [email p o ec ed]; E angelos Alexiou,Xiaomi Technology, The Hague, The Ne he lands;
e-mail: [email p o ec ed]; Jack Jansen,Cen um Wiskunde & In o ma ica, Ams e dam, The Ne he lands; e-mail:
[email p o ec ed]; Pablo Cesa ,Cen um Wiskunde & In o ma ica, Ams e dam, TU Del , Del , The Ne he lands; e-mail:
[email p o ec ed].
This wo k is licensed unde C ea i e Commons A ibu ion In e na ional 4.0.
© 2025 Copy igh held by he owne /au ho (s).
ACM 1551-6865/2025/8-ART230
h ps://doi.o g/10.1145/3731759
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:2 X. Zhou e al.
Addi ional Key Wo ds and Ph ases: Volume ic ideo, Dynamic poin cloud, Visual saliency, Visual a en ion,
Subjec i e quali y assessmen , Objec i e quali y me ics, Eye acking, 6DoF
ACM Re e ence o ma :
Xuemei Zhou, I ene Viola, E angelos Alexiou, Jack Jansen, and Pablo Cesa . 2025. Subjec i e and Objec i e
Quali y Assessmen o Dynamic Poin Cloud wi h Visual A en ion in 6 DoF. ACM T ans. Mul imedia Compu .
Commun. Appl. 21, 8, A icle 230 (Augus 2025), 24 pages.
h ps://doi.o g/10.1145/3731759
1 In oduc ion
Volume ic ideo has become a ailable o ep esen eal-wo ld objec s due o he apid de elopmen
o cap u e de ices, ansmission echnologies, and compu a ional capabili ies. Poin cloud has
eme ged as one o he mos popula o ma s o olume ic ideo ep esen a ion. Speci ically,
Dynamic Poin Clouds (DPCs) can be used o au omo i e/ obo ic na iga ion [74], medical
imaging [9], and i ual ideo con e encing [30,61], among o he applica ions. A poin cloud
ame can be de ined as a se o poin s in space ep esen ed in a 3D coo dina e sys em. Each poin
cloud ame equi es a la ge numbe o poin s o ai h ully ep esen he con en and achie e a
good Quali y o Expe ience (QoE). A DPC is essen ially a sequence o indi idual poin cloud
ames played in succession. The e o e, e ec i e comp ession is essen ial be o e ansmission,
s o age, ende ing, and display. Quali y deg ada ion will be ine i ably in oduced du ing his end-
o-end pipeline, which de e io a es he isual quali y and a ec s human pe cep ion. Explo ing he
dis o ion cha ac e is ics o DPCs and e ec i ely measu ing hem is a challenge in bo h subjec i e
and objec i e quali y assessmen [3].
Subjec i e quali y assessmen leads o g ound- u h a ings o isual impai men s ha appea
in a s imulus. Subjec i e quali y assessmen o DPC has been explo ed in passi e desk op iewing
condi ions [70,71] o in imme si e en i onmen s wi h use s consuming he con en s h ough a
Head-Moun ed Display (HMD) unde 6 Deg ees o F eedom (DoF) [53,63]. In he la e case,
in o ma ion abou use s’ mo emen s can be cap u ed in addi ion o subjec i e quali y a ings,
o unde s and how use s na iga e and obse e objec s in Vi ual Reali y (VR) space. A mo e
accu a e ep esen a ion o he use ’s consump ion is gi en by gaze da a, which highligh he speci ic
a eas o con en being iewed wi h ocused a en ion. This in o ma ion aids in he c ea ion o isual
a en ion maps. Inco po a ing isual a en ion in o quali y assessmen has demons a ed po en ial
imp o emen o p edic ing he isual quali y o 2D/3D image/ ideo [33,73]. None heless, isual
a en ion o DPCs is unexplo ed, hus hinde ing he u iliza ion o i s ou comes in aiding isual
quali y assessmen .
A summa y o exis ing subjec i e quali y assessmen and isual a en ion da ase s o poin
clouds is shown in Table 1. Mos o he s udies in he li e a u e in ol ing DPCs a e conduc ed in
desk op se ups wi h 2D moni o s, whe e he DPCs a e p e- eco ded and he playback is conduc ed
using con en ional ideo so wa e [68,70,71]. Howe e , such desk op se ups limi use in e ac ions
(i any) o mouse and keyboa d, es ic ing use eedom. In con as , in imme si e HMD-based
se ups, a mo e na u al consump ion wi h 6 DoF h ough use mo emen s is enabled, allowing o a
mo e ealis ic ep esen a ion o he en i e DPC. The echnical challenges aced wi h HMD-based
se ups (e.g., eal- ime playback o DPCs [53]), hough, lead o expe imen al sessions ha ypically
in ol e a smalle numbe o DPCs (20), usually s a ic o wi h sho e ime du a ion (5 seconds).
Main aining he accu acy o he eye- acke in 6 DoF is ye ano he challenge [40]. Due o such
cons ain s, no isual a en ion da ase speci ically designed o DPCs consumed by HMD-based
se ups in 6 DoF has been eleased so a ; exis ing esea ch has p ima ily explo ed he a en ion o
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:3
Table 1. Publicly A ailable Subjec i e Quali y Assessmen and Visual A en ion Da ase s o Poin Clouds
Da ase Type Deg ada ion S imuli Time Display In e ac ion Opinion Sco e Visual A en ion
VsenseVVDB [70] Dynamic Down-sampling, V-PCC 32 6.6 s 2D moni o 737
VsenseVVDB2 [71] Dynamic Mesh: D aco+JPEG
Poin Clouds: G-PCC, V-PCC
28
136 10 s 2D moni o 737
Owlii [68] Dynamic Mesh: TFAN, FFmpeg
Poin Clouds: V-PCC, FFmpeg 20 20 s 2D moni o 737
Ma ie e al. [37] Dynamic Posi ion, Tex u e coo dina e
HEVC, T iangle holes 176 10 s 2D moni o 737
VOLVQAD [12] Dynamic V-PCC 376 10 s 2D moni o 737
Sub amanyam e al. [53] Dynamic CWI-PCL, V-PCC 72 5 s HMD 3 3 7
ComPEQ-MR [42] Dynamic V-PCC, G-PCC 52 10 s Augmen ed eali y 3 3 3
ViA PCVR [5] S a ic Only e e ence 8 - HMD 373
QAVA-DPC (Ou s) Dynamic V-PCC, G-PCC, CWI-PCL 50 10 s HMD 3 3 3
Fig. 1. Fixa ion maps o dance sequence wi h uni o m empo al sampling e e y 30 ames.
s a ic poin clouds [5], con ining he scope o a ew undis o ed con en s. Mo e gene ally, he e is
cu en ly a lack o s udies ha connec isual a en ion and isual quali y speci ically o DPC.
Visual a en ion plays a c ucial ole in a ious ision asks, such as segmen a ion, localiza ion,
and egis a ion [22]. No ably, le e aging isual a en ion maps o weigh quali y maps has shown
imp o emen s in pe cep ual quali y p edic ion [35]. By connec ing isual a en ion and isual
quali y o DPCs, quali y alloca ion be ween salien egions and su ounding a eas, saliency-awa e
comp ession and s eaming, and saliency-aided objec i e quali y me ics can be u he in es iga ed
and op imized.
In ou p e ious pape [75], we c ea ed an eye- acking-based Quali y Assessmen and Visual
A en ion da ase o DPCs (QAVA-DPC). The da ase includes di e se con en and a ious
comp ession dis o ions using he MPEG s anda d codecs: Video-Based Poin Cloud Com-
p ession (V-PCC),Geome y-Based Poin Cloud Comp ession (G-PCC), and he MPEG
e e ence codec, e e ed o he e as CWI-PCL. Wi h his wo k, we aim o ex end ou p e ious e o s
by conduc ing an in-dep h analysis o semi-s uc u ed in e iews conduc ed a e he subjec i e
expe imen s and by benchma king objec i e poin cloud quali y assessmen me ics wi h collec ed
isual a en ion da a. The added alue o associa ed isual a en ion maps and esul s om he
semi-s uc u ed in e iews can he eby enhance ou unde s anding o human beha io wi hin 6
DoF en i onmen s, ul ima ely con ibu ing o he op imiza ion o QoE. Ou con ibu ions can be
summa ized as ollows:
—
We p opose a new da ase , namely, QAVA-DPC, which con ains i e e e ence DPCs; each
DPC is encoded by h ee codecs, wi h each codec con igu ed a h ee dis o ion le els. Fixa ion
maps a e cons uc ed, collec ed, and p esen ed o bo h he e e ence and dis o ed sequences
as hea maps o e laid on op o he s imuli ames. To he bes o ou knowledge, his is he
i s ime connec ing isual a en ion and isual quali y o DPCs in VR.
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:4 X. Zhou e al.
—
We benchma k s a e-o - he-a me ics, ini ially in ended o s a ic poin clouds, along wi h
wo empo al pooling me hods on he QAVA-DPC da ase . In addi ion, we alida e he pe o -
mance o hese me ics using g ound- u h isual a en ion maps.
—
We elease all aw da a, con aining he opinion sco es and gaze samples collec ed in ou
s udy, alongside he so wa e used o pe o m he expe imen , and he sc ip s used o expo
isual a en ion maps, a he ollowing link: h ps://gi hub.com/cwi-dis/ISMAR_Poin Cloud_
EyeT acking.
2 Rela ed Wo k
2.1 Subjec i e Quali y Assessmen o DPC
Whe eas subjec i e quali y assessmen o s a ic poin clouds has been explo ed in mo e de ail in he
li e a u e [4], analogous esea ch on DPCs is s ill a sophis ica ed and challenging p oblem, owing
o nume ous ac o s such as he e alua ion me hodology, ende ing me hod, display equipmen ,
and so o h. Subjec i e quali y sco es, such as Mean Opinion Sco e (MOS) o Di e en ial MOS
(DMOS), a e commonly used o quan i y he subjec i e pe cep ion o isual a i ac s. Ze man
e al. [70] conduc a subjec i e expe imen on wo DPCs om he VsenseVVDB da ase ha a e
dis o ed using V-PCC comp ession [49]. They a gue ha ce ain geome ic dis o ion me ics a e
incong uen wi h he expec ed quali y. Hoo e al. in es iga e how and o wha ex en a ious
aspec s impac he use ’s QoE, ia ex ensi e subjec i e e alua ion o olume ic 6 DoF s eaming
[58]. Meku ia e al. e alua e he subjec i e quali y o he CWI-PCL codec pe o mance using a
ealis ic 3D ele-imme si e sys em in a i ual oom scena io, in which use s a e ep esen ed and
in e ac as 3D a a a s and/o 3D poin clouds [38]. The subjec i e s udy shows ha in oduced
p edic ion dis o ions a e negligible compa ed wi h he o iginal econs uc ed poin clouds. Cao
e al. [8] s udy he pe cep ual quali y o comp essed 3D sequences, o bo h poin cloud comp ession
and mesh-based comp ession. They explo e he impac o bi a e and obse a ion dis ance on
pe cep ual quali y. Cox e al. [12] p esen VOLVQAD, a olume ic ideo quali y assessmen da ase
wi h 376 ideo sequences. The olume ic ideo sequences a e i s encoded wi h MPEG V-PCC
using 4 di e en a a a models and 16 quali y a ia ions, and hen ende ed in o es ideos o
quali y assessmen using 2 di e en backg ound colo s and 16 di e en quali y swi ching pa e ns
wi h a 2D display. Howe e , hese expe imen s a e pe o med in a desk op se up. Viola e al. [63]
compa e wo di e en VR iewing condi ions enabling 3/6 DoF, along wi h a desk op se up, o
unde s and how in e ac ion in he i ual space a ec s he pe cep ion o quali y. Resul s show
no s a is ical di e ence be ween sco es gi en in a desk op and VR se up; howe e , quali a i e
esul s highligh ed he added alue o in e ac i e e alua ions. One limi a ion o he s udy lies in he
ime du a ion (5 seconds) o he sequences used o he e alua ion, as he au ho s use 150 ames.
Sub amanyam e al. [54] e alua e he pe o mance o se e al adap i e s eaming solu ions in an
in e ac i e VR expe imen . They compa e he pe o mance o V-PCC wi h espec o CWI-PCL,
using a ious adap i e s eaming s a egies. Quan i a i e subjec i e esul s and quali a i e insigh s
indica e ha V-PCC has a mo e a o able pe o mance han he CWI-PCL, especially a low bi
a es. Damme e al. [56] conduc ed an in-dep h subjec i e s udy on he impac o con e ing poin
clouds o meshes wi h a ying-quali y ep esen a ions. Addi ionally, while end-use s demons a e
awa eness o quali y swi ches, he e ec on hei pe cep ion emains limi ed. Gu ié ez e al.
[20] p esen a subjec i e s udy on DPCs conside ing di e en comp ession a es using he MPEG
s anda d V-PCC. Resul s on use s’ explo a ion beha io show no signi ican di e ences when
isualizing poin clouds wi h di e en quali ies, no changes in he beha io du ing he es session,
and no co ela ion be ween explo a ion ac i i y and quali y assessmen s. Nguyen e al. [42] p o ide
an open sou ce comp essed poin cloud da ase wi h eye- acking da a and quali y assessmen
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:5
sco e in mixed eali y wi h Hololens 2, including 4 DPCs. Eye- acking da a and opinion sco es
a e collec ed unde di e en expe imen al se ings. While nume ous s udies ha e explo ed he
subjec i e quali y assessmen o DPCs, he e is a gap in esea ch ocusing on isual a en ion and
quali y assessmen in VR wi h 6 DoF.
2.2 Objec i e Quali y Assessmen o DPC
Objec i e quali y assessmen o 2D/3D ideo has achie ed ema kable p og ess in ecen yea s.
Howe e , ew speci ic objec i e quali y me ics ha e been designed o DPCs so a . Ak e al. [1]
explo e he possibili y o empo al sub-sampling o he con en unde e alua ion o objec i e
quali y e alua ion wi hou sac i icing he co ela ion wi h he subjec i e opinion. Thi y di e en
objec i e quali y me ics a e es ed on he VsenseVVDB2 da ase , combined wi h empo al sub-
sampling and empo al pooling me hods. Resul s show ha he pe o mance o objec i e me ics
o poin cloud comp ession is minimally a ec ed by he empo al sub-sampling a e. F ei as
e al. [18] in es iga e he added alue o inco po a ing empo al pooling in o he DPC’s quali y
assessmen model using me ics designed o s a ic poin clouds. They ind ha he pe o mance o
empo al pooling is consis en ly be e when a empo al a ia ion model [43] is used. The same
au ho s in es iga e he sui abili y o geome ic-awa e ex u e desc ip o s o blindly assess he
quali y o colo ed DPCs [19] on op o he same empo al pooling s a egy, leading o simila
conclusions. Yang e al. [69] conduc a subjec i e use s udy o unde s and he e ec i eness o
di e en pe cep ual quali y me ics o olume ic ideo and design an objec i e me ic called
Volu-FMAF. Volu-FMAF combines poin -based and pixel-based me ics wi h iewpoin - ela ed
ea u es. They u he p opose a dis o ion-awa e ende ed image supe - esolu ion ne wo k in a
olume ic ideo s eaming amewo k, which exploi s he insigh s ob ained om hei use s udy.
Damme e al. [57] p esen a ho ough co ela ion analysis o bo h ull- e e ence and no- e e ence
objec i e me ics o subjec i e MOS wi h a double pu pose. Addi ionally, hey in es iga e how
egion o in e es selec ion and weigh ing p ocedu es impac he accu acy o enhance i u he . The
s udy shows ha he classical ideo quali y me ic VMAF [32] is e y well sui ed as an objec i e
benchma k o olume ic media s eaming in e ms o co ela ion o subjec i e sco es, and a
combina ion o no- e e ence ea u es could p o ide a good eal- ime assessmen . Fan e al. [16]
p opose a deep-lea ning-based no- e e ence olume ic ideo quali y assessmen me hod based
on mul i- iew lea ning. They i s p ojec olume ic ideos o 2D ideo sequences om a ious
iewpoin s. Then, ResNe 3D is u ilized o ex ac quali y-awa e ea u es, and a quali y eg ession
module is designed o use he ea u es lea ned om he mul iple iewpoin s and join ly eg ess
hem in o quali y sco es. Ma ie e al. [37] benchma k and calib a e se e al objec i e quali y
me ics on a challenging olume ic ideo da ase ep esen ed as ex u ed meshes. They e alua e
wo model-based app oaches—MPEG PCC and PCQM—by con e ing meshes in o poin clouds
using a de ised mesh su ace sampling me hod, as well as an image-based app oach (IBSM), o
which hey in oduce wo new ea u es speci ically designed o de ec holes and empo al de ec s.
Fo each me ic, he op imal selec ion and combina ion o ea u es a e de e mined by logis ic
eg ession h ough c oss- alida ion. The pe o mance analysis, combined wi h MPEG expe s’
equi emen s, leads o ecommenda ions on he ea u es o highes impo ance h ough lea ned
ea u e weigh s, such as empo al pooling, in eg a ing an a en ion model.
2.3 Visual A en ion-Based Objec i e Quali y Assessmen
Recen li e a u e in eye- acking-based isual saliency o imme si e con en s has mainly ocused
on ask- ee expe imen s o ga he isual a en ion maps [45,51]; no s udy has been conduc ed
o link isual a en ion o isual quali y assessmen o olume ic ideos. The li e a u e sugges s
ha isual a en ion migh be bene icial o unde s anding he p ocess o pe cep ion o isual
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:6 X. Zhou e al.
quali y o 2D images/ ideos; in ac , di e en me ics o Image Quali y Assessmen (IQA) ha e
been ex ended wi h a compu a ional model o isual a en ion [33], bu he esul ing gain on he
me ics’ pe o mance is so a unclea . To be e unde s and he added alue o including isual
a en ion in he design o objec i e me ics o 2D images, some wo ks in he li e a u e ha e aken
ad an age o eco ded isual a en ion da a. Lin and Heynde ickx [34] pe o m wo eye- acking
expe imen s: one wi h a ee-looking ask and one wi h a quali y assessmen ask. They ound a
endency ha adding saliency o a me ic yields a la ge amoun o gain in pe o mance. The ex en
o he pe o mance gain ends o depend on he speci ic objec i e me ic and he image con en . In
addi ion, he gain is small o objec i e me ics ha al eady show a high co ela ion wi h pe cei ed
quali y o a gi en dis o ion ype. Zhang and Liu [73] p opose a new me hodology o elimina e
he inhe en bias due o he in ol emen o s imulus epe i ion. The e ined me hodology esul s
in a new eye- acking da ase wi h a la ge deg ee o s imulus a iabili y. Based on g ound- u h
labeling, he s a is ical e alua ion shows ha he isual a en ion in o ma ion o bo h he e e ence
and he dis o ed scene is bene icial o IQA me ics, bu he la e ends o u he boos he
e ec i eness o in eg a ing a en ion in IQA me ics. Jin e al. [27] u ilize an eye- acke o c ea e
o ea ion-comp essed VR da ase s and e alua e bo h he o ea ed and non- o ea ed objec i e
image/ ideo quali y assessmen algo i hms.
To be e unde s and whe he he indings ega ding isual saliency and quali y assessmen on
2D images/ ideos can hold o olume ic ideos, ad-hoc da ase s, and benchma king alida ion
o objec i e me ics ha combine he wo aspec s a e needed. Tha is he esea ch gap we aim o
ill wi h his a icle.
In his s udy, ou objec i e is o en ich he exis ing li e a u e by conduc ing a subjec i e expe i-
men ha compa es he isual quali y o se e al s a e-o - he-a comp ession echniques o DPC.
This expe imen is in e ac i e, using an HMD-based VR ende ing o 10-second DPC sequences
om a ious da ase s, a me hodology no p e iously explo ed in conjunc ion wi h eye- acking in
he li e a u e. In addi ion, we del e in o he analysis o in e iews conduc ed du ing he subjec i e
expe imen o gain deepe insigh s in o use expe iences and he ac o s ha in luence he pe cep-
ual quali y o DPCs. Fu he mo e, we alida e widely used objec i e me ics and pooling me hods
epo ed in he li e a u e o DPC objec i e quali y assessmen .
3 QAVA-DPC Cons uc ion
3.1 Con en Selec ion
Fo he c ea ion o he da ase , we selec ed i e DPCs om h ee public da ase s, namely Vsen-
seVVDB2 [71], 8i [14], and Owlii [68]. To show he di e si y o DPCs, we conside ed he Spa ial
In o ma ion (SI) and Tempo al In o ma ion (TI) o each con en [24]. We p ojec ed he sou ce
poin cloud in o ou iews, which a e he le , igh , on , and back iew, o i s bounding box o
apply SI and TI sepa a ely [67], hen ob ain he maximum alue among he ou iews o e all he
i s 300 ames as he inal SI/TI o one sequence. Thei dis ibu ion o all DPCs can be seen in
Figu e 2. The dispe sion in SI (ho izon al axes)/TI ( e ical axes) shows he di e si y o ou con en s
in he spa ial/ empo al domain. We inally chose dance ,exe cise,long d ess,Ra a2, and soldie as
he con en s o ou da ase .
3.2 S imuli P ocessing
P io o he subjec i e expe imen on DPCs, speci ic p ocedu es, such as p e-p ocessing, encoding,
and ende ing, a e necessa y due o codec implemen a ions, wi h he goal o minimizing addi ional
in luencing ac o s.
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:7
Fig. 2. Dis ibu ion o SI and TI o 12 sou ce DPCs om h ee da ase s, he colo alue is compu ed by
p(𝑆𝐼2+𝑇𝐼2). SI, Spa ial In o ma ion; TI, Tempo al In o ma ion.
3.2.1 P e-P ocessing. The sequences men ioned abo e a e selec ed om di e en da ase s, which
means ha he esolu ion, posi ion, and o ien a ions a y. To c ea e a ealis ic ele-imme si e
scena io, he DPCs should be displayed as li e-size models. To do so, we no malized he DPCs o a
simila bounding box. Fo dance and exe cise, he geome y p ecision is oxelized om 11 o 10 bi s
o ensu e consis ency wi h he o he con en s. The sou ce models we e p ocessed wi h o a ion,
ansla ion, and scaling. Addi ionally, since he V-PCC encode ope a es wi h in ege alues, he
coo dina es o all DPCs we e ounded be o e V-PCC comp ession. CWI-PCL encode has speci ic
equi emen s o he esolu ion o DPCs, so be o e CWI-PCL comp ession, he coo dina es wen
h ough a scaling ope a ion.
3.2.2 Encoding. Dis o ed e sions we e gene a ed using he s a e-o - he-a MPEG PCC e -
e ence so wa e Tes Model Ca ego y 2 Ve sion 18 (
𝑇𝑀𝐶2
V-18.0) and Ca ego y 1&3 Ve sion 14
(
𝑇𝑀𝐶13
V-14.0) [49], which implemen V-PCC and G-PCC, espec i ely. We also adop he CWI-PCL
[38] codec as a compa ison, which se ed as a base e e ence so wa e pla o m in MPEG. G-PCC
was o iginally p oposed o comp ess s a ic poin clouds, al hough mo e ecen e sions a e capa-
ble o handling DPCs; i ope a es in he 3D, poin cloud domain. V-PCC was de eloped o DPC
comp ession; i ope a es in he 2D, ideo domain. CWI-PCL was designed as a ligh weigh codec
ha complies wi h eal- ime equi emen s; he geome y is encoded in he 3D and he a ibu es in
he 2D domain. To compa e hem in a ai way, we se he G-PCC encode o use Region-Ap i e
Hie a chical T ans o m (RAHT) and Oc ee o comp essing colo a ibu es and geome y,
espec i ely; he V-PCC encode o use All In a (AI) mode, which applies in a-p edic ion o all
ames; and he CWI-PCL encode o use in a-p edic ion in all ames, wi h oc ee subdi ision
and JPEG codec o comp ess geome y and colo a ibu es, espec i ely.
To de ine he con igu a ion pa ame e s o he encode s, he MPEG Common Tes Condi ions
(CTC) [52] a e ollowed. Mo eo e , we selec h ee dis o ion le els ha ep esen compa able low,
medium, and high-quali y anges o each encode . Speci ically, o G-PCC, we selec R02, R04, and
R05 om he MPEG CTC, which a e ealized by adjus ing he posi ionQuan iza ionScale and he QP
pa ame e s. Fo V-PCC, we selec R01, R03, and R05 by adjus ing he geome y QP, he a ibu e
QP, and he occupancyP ecision pa ame e s. Fo CWI-PCL, we choose h ee combina ions o oc ee
dep h wi h JPEG QP pa ame e s o ma ch a simila quali y ange, by looping o e oc ee dep h
om 7 o 9 and JPEG QP om 25 o 95 (s ep size
=
10). When es ed on he da ase , he speci ied
pa ame e se ings o he h ee codecs yielded subjec i ely simila quali y anges. The speci ic
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:8 X. Zhou e al.
Table 2. Pa ame e Se s o he Selec ed Encode s
Encode s Dis o ion Le el
G-PCC
(Oc ee-RAHT)
R02
(0.125, 46)
R04
(0.5, 34)
R05
(0.75, 28)
V-PCC (AI) R01
(32,42, 4)
R03
(24, 32, 4)
R05
(16, 22, 2)
CWI-PCL R01
(7, 25)
R02
(8, 95)
R03
(9, 95)
Fig. 3. Schema ic diag am wi h he ha dwa e and so wa e modules oge he wi h hei in e -dependencies.
pa ame e se ings a e shown in Table 2. Each e e ence DPC is comp essed using 3 encode s, and
each encode has 3 dis o ion le els, o a o al o 45 dis o ed DPCs.
3.2.3 Rende ing. Rende ing is he p ocess o p oducing a isual ep esen a ion ha can be
consumed by use s using an a ailable display. In he case o poin clouds, di e en ende ing
me hods may ha e a signi ican impac on pe cei ed quali y [26]. In ou expe imen , we chose o
ende he poin clouds wi hou any addi ional p ocessing (e.g., su ace econs uc ion), di ec ly
using he poin cloud da a (poin -based).
Ou expe imen so wa e is de eloped in Uni y ( e sion 2021.3.10. 1), exploi ing he S eamVR
plugin ( e sion 1.24.7) o connec wi h VR headse s and con olle s. CWI Poin Cloud (CWIPC)-
suppo ed uni y package ( e sion 0.10.0) helps us impo he DPCs and play hem back inside Uni y
[46]. A high-le el diag am indica ing he ha dwa e/so wa e dependencies is p o ided in Figu e 3.
No ably, a la ge size o DPC ile migh ake up oo much memo y and cause a sys em hang.
We con e ed he DPC da a in o he CWIPC-suppo ed poin cloud playback o ma o enhance
so wa e s abili y wi hou in oducing addi ional dis o ion. To ensu e smoo h playback o DPCs,
we ook ad an age o he Uni y Co ou ine scheme o p eload each DPC in o memo y be o e he
use swi ches o he nex DPC. I should be no ed ha o each DPC sequence, we only chose he
i s 300 ames om he sou ce model. The ame a e o playback was se o 30 ames pe second,
hence, each DPC sequence las s o 10 seconds.
Fo he same s imuli, bo h he e e ence and dis o ed e sions we e made o appea wa e igh
by adjus ing he poin size based on he a e age dis ance o each poin ’s 10 nea es neighbo s
ac oss he en i e poin cloud [53]. Wi hin a DPC, we u ilized he same poin size o all ames.
All he poin clouds we e escaled o a simila size, a ound 1.8 m in heigh , o mimic a ealis ic
ele-imme si e scena io. The VR scene was illumina ed by a i ual lamp on he ceiling, cen alized
abo e he models. The lamp was se as an a ea ligh wi h in ensi y alues o 2 in Uni y o simula e
o dina y ligh ing in a oom. We used HTC Vi e P o Eye de ices wi h eye- acking capabili ies
and Vi e hand con olle s o pa icipan s o in e ac in ou expe imen . To de elop eye- acking
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:9
Fig. 4. Playback scene.
applica ions o he Vi e P o Eye, we used he na i e HTC Vi e SRanipal SDK. The sampling
equency (binocula ) o he eye acke was 120 HZ.
3.3 Expe imen al P ocedu e
Since he e is no speci ic ecommenda ion o designing subjec i e quali y assessmen expe i-
men s o DPCs in VR, we ha e made an e o o adhe e o exis ing ITU ecommenda ions on
images/ ideos [21,23,25] o es ablish an app op ia e assessmen me hodology o DPCs. In ou
subjec i e s udy, we op ed o he Absolu e Ca ego y Ra ing wi h Hidden Re e ence (HR) using
i e-le el a ings (1-Bad, 2-Poo , 3-Fai , 4-Good, and 5-Excellen ). Each ime, only a single DPC was
shown o he obse e ; es ma e ials included impai ed DPCs wi h andomly inse ed in ac HRs,
ep esen ed as any o he es s imulus. To a oid he e ec s o con ex ual o memo y compa isons,
we andomly gene a ed a playlis o each subjec , and ca e was gi en o a oid displaying he same
con en consecu i ely.
Be o e he expe imen , he isual acui y and colo ision o e e y subjec we e es ed using
Snellen [17] and Ishiha a [11] cha s. Each subjec was in o med in ad ance abou he manne and
pu pose o he s udy as pa o he in o med consen p ocedu e. A he beginning o he session,
he in e -pupilla y dis ance was measu ed and he headse was adjus ed by he subjec acco dingly.
Then, a aining session was conduc ed o help amilia ize he subjec s wi h he sys em, including
he con olle s and he naming o each bu on o communica e mo e easily. One aining sequence,
namely loo , was used, which was no included in he da ase . The quali y ange o loo was simila
o he quali y ange o he es ideos, gi ing he subjec s a sense o wha hey would see in he
o mal sessions. The subjec s always s a ed a he same loca ion, which is 1.5 m away om he
cen e o he i ual oom, bu could mo e eely om he e onwa d.
A DPC was loca ed in he cen e o he i ual oom, as shown in Figu e 4, and each DPC
was andomly o a ed be ween
[0◦,360◦]
o a oid bias. Du ing he expe imen , he subjec s we e
ins uc ed o iew each model ca e ully in he VR en i onmen , by mo ing eely du ing he
playback o each DPC. The subjec s we e also equi ed o s and s ill while doing he calib a ion
and e o p o iling. A ixed dis ance was se be ween he subjec s and he e o p o iling scene,
which was a ci cle composed o nine eye-ball ma ke s.
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:16 X. Zhou e al.
Fig. 11. Fixa ion map on he hai and heel o long d ess.
Fig. 12. The e e enced and dis o ed e sions o poin cloud long d ess ( ame 128) wi h co esponding isual
a en ion maps based on he p oposed p ocessing p o ocol. (a) The e e ence e sion. (b)–(j) The dis o ed
e sions o long d ess om low o high quali y. Speci ically, (b)–(d): V-PCC, (e)–(g): G-PCC, (h)–(j): CWI-PCL.
A i e-pa ame ic logis ic eg ession is adop ed o i he ela ionship be ween he subjec i e sco es
and he objec i e sco es be o e calcula ing he PLCC and RMSE [6].
Pe o mance E alua ion o Objec i e Quali y Me ics wi hou Visual A en ion. Wi h he subjec i e
sco es collec ed in ou expe imen s, we conduc an e alua ion and compa ison o exis ing objec i e
me ics o he ask o DPC quali y assessmen . Only poin -based me ics a e conside ed. We
chose me ics adop ed by he MPEG g oup, namely, poin - o-poin and poin - o-plane wi h MSE
and Hausdo dis ances, wi h and wi hou using Peak Signal o Noise Ra io (PSNR), and he
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:17
weigh ed a e age colo di e ences o Y, U, and V channels in e ms o PSNR, which is de ined as
[55]:
𝑃𝑆𝑁𝑅𝑌𝑈𝑉 =
6·𝑃𝑆𝑁𝑅𝑌+𝑃𝑆𝑁𝑅𝑈+𝑃𝑆𝑁𝑅𝑉
8.(3)
Fu he mo e, we choose ano he ou s a e-o - he-a me ics, namely, his Y [62], poin SSIM [2],
PCM
_
RR [60], and PCQM[39]. Since he me ics a e o iginally designed o s a ic con en s and do
no explici ly conside he empo al aspec , we apply he me ics o each ame and hen pool all
he 300 ames’ sco es as he inal sco e o one DPC sequence. Based on he indings o [19], we
choose he mean as he baseline and he a ia ion model o [43] as he empo al pooling me hod o
ob ain he objec i e sco e o each DPC sequence.
Pe o mance E alua ion o Objec i e Quali y Me ics wi h Visual A en ion. Since he isual saliency
is poin -based, we in eg a ed he impo ance weigh wi h he quali y sco e compu ed by he exis ing
me ic o each poin pe ame, and adop ed he same pooling s a egy, o ob ain he inal quali y
sco e. The impo ance weigh is no malized be ween 0 and 1 based on he compu a ion esul in
Equa ion (2). We adop ed wo ways o weigh ing he poin -based me ics wi h isual saliency. The
i s conside s only he salien egion and excludes he unsalien egion, which is de ined as:
𝑄1
𝑣=𝑀·𝑉𝑆𝑓,(4)
whe e
𝑄1
𝑣
is he quali y sco e wi h only he salien egion o one ame belonging o a DPC, and
𝑀
is he quali y sco e o he co esponding ame om he poin -based me ics. The second me hod
e ains he unsalien egion bu assigns ela i ely highe weigh s o he salien egion, e med as
no malized saliency, which is de ined as:
𝑄2
𝑣=𝑀· (𝑉𝑆𝑓+1),(5)
whe e 𝑄2
𝑣is he quali y sco e wi h he no malized saliency o one ame belonging o a DPC.
The esul s o he pe o mance indices o he 13 objec i e me ics, wi h mean and a ia ion as
he empo al pooling me hods, a e p esen ed in Tables 3and 4. Consis en wi h indings in [1],
we no e ha al e ing he empo al pooling me hod does no signi ican ly impac high-pe o ming
quali y me ics (PLCC highe han 0.5). In he o iginal implemen a ion, PCM
_
RR achie es he
highes PLCC/SRCC pe o mance wi h a e age pooling, p2poin _Hausd o demons a es he bes
pe o mance when only conside ing he salien egion wi h a e age pooling, his _Y achie es he
highes SRCC pe o mance a e applying no malized salien weigh ing, and poin SSIM achie es
he bes PLCC/RMSE, and PCQM achie es he bes SRCC a e no malized saliency. The signi ican
pe o mance inc ease o p2poin
_
Hausd o me ic may be a ibu ed o he exclusion o he non-
salien egion, which helps mi iga e he sensi i i y o ou lie s. This e inemen ensu es ha only
e o s wi hin he salien egion a e e ained. Gene ally, se e al s udies epo ha mode n ideo
quali y assessmen models expe ience only sligh imp o emen s om inco po a ing saliency [64].
Howe e , o DPC quali y assessmen , me ics ha ocus exclusi ely on salien egions show a
decline in pe o mance. Con e sely, me ics using a no malized weigh ing s a egy achie e simila
esul s o he o iginal implemen a ion, wi h pe o mance also depending on he empo al pooling
me hods employed.
4.4 Quali a i e Resul s
Thi y- wo in e iew audio eco dings we e ansc ibed in o ex s and coded using Do e ail,
1
wi h
eigh pa icipan s declining o pa icipa e. Following Magui e’s guideline on hema ic analysis [36],
we ini ially e iewed and labeled he ex , o ganized hese labels in o hemes, and subsequen ly
1h ps://do e ail.com/
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:18 X. Zhou e al.
Table 3. Pe o mance E alua ion o S a e-o - he-A Quali y Me ics on QAVA-DPC wi h A e age Pooling
Me ics PLCC SRCC RMSE
OI SPO NSW OI SPO NSW OI SPO NSW
p2poin _MSE 0.782 0.613 0.781 0.738 0.544 0.738 0.728 0.902 0.730
p2poin _MSE_PSNR 0.513 0.580 0.513 0.465 0.555 0.465 1.002 0.929 1.003
p2poin _Hausd o 0.200 0.854 0.434 0.240 0.687 0.323 1.144 0.594 1.052
p2poin _Hausd o _PNSR 0.521 0.661 0.364 0.146 0.641 0.204 0.997 0.856 1.088
p2plane_MSE 0.710 0.571 0.710 0.690 0.546 0.689 0.822 0.937 0.823
p2plane_MSE_PSNR 0.628 0.597 0.484 0.483 0.571 0.483 0.908 0.915 1.026
p2plane_Hausd o 0.207 0.767 0.428 0.257 0.634 0.281 1.142 0.732 1.055
p2plane_Hausd o _PNSR 0.514 0.699 0.532 0.163 0.617 0.183 1.005 0.816 0.994
YUV_PSNR 0.661 0.606 0.662 0.654 0.561 0.652 0.876 0.907 0.875
his _Y 0.840 0.204 0.827 0.820 0.086 0.781 0.633 1.143 0.657
PCM_RR 0.844 0.058 0.444 0.822 0.046 0.398 0.626 1.166 1.046
poin SSIM 0.836 0.652 0.836 0.772 0.649 0.771 0.641 0.885 0.640
PCQM 0.813 0.334 0.833 0.758 0.256 0.815 0.681 1.101 0.646
Columns ep esen O iginal Implemen a ion (OI), Salien -Pa -Only (SPO), and No malized Saliency Weigh ing (NSW). The
bes pe o mance o PLCC/SRCC/RMSE is highligh ed in bold and ma ked wi h ed/blue/o ange colo , espec i ely.
Table 4. Pe o mance E alua ion o S a e-o - he-A Quali y Me ics on QAVA-DPC wi h Va ia ion Pooling
Me ics PLCC SRCC RMSE
OI SPO NSW OI SPO NSW OI SPO NSW
p2poin _MSE 0.769 0.604 0.769 0.731 0.536 0.730 0.747 0.909 0.751
p2poin _MSE_PSNR 0.267 0.108 0.267 0.135 0.180 0.135 1.125 1.134 1.125
p2poin _Hausd o 0.527 0.593 0.515 0.251 0.536 0.223 0.993 0.919 1.001
p2poin _Hausd o _PNSR 0.248 0.114 0.319 0.232 0.069 0.286 1.131 1.133 1.107
p2plane_MSE 0.662 0.570 0.662 0.621 0.539 0.625 0.875 0.937 0.875
p2plane_MSE_PSNR 0.210 0.103 0.210 0.119 0.134 0.119 1.141 1.134 1.142
p2plane_Hausd o 0.453 0.605 0.501 0.163 0.556 0.217 1.041 0.909 1.010
p2plane_Hausd o _PNSR 0.361 0.054 0.327 0.242 0.005 0.299 1.089 1.139 1.103
YUV_PSNR 0.643 0.601 0.643 0.625 0.583 0.625 0.894 0.912 0.894
his _Y 0.735 0.249 0.669 0.727 0.054 0.617 0.792 1.133 0.868
PCM_RR 0.552 0.393 0.293 0.485 0.216 0.020 0.974 1.074 1.146
poin SSIM 0.705 0.594 0.707 0.710 0.539 0.714 0.828 0.940 0.825
PCQM 0.807 0.357 0.732 0.738 0.256 0.552 0.690 1.091 0.796
Columns ep esen OI, SPO, and NSW. The bes pe o mance o PLCC/SRCC/RMSE is highligh ed in bold and ma ked wi h
ed/blue/o ange colo , espec i ely.
con ened o es ablish he connec ion be ween pe cep ual quali y and isual a en ion du ing he
subjec i e es . Each pa icipan is deno ed as P1–P32, wi h he numbe o pa icipan s concu ing
wi h each s a emen indica ed in pa en heses. The quali a i e esul s o his hema ic analysis a e
p esen ed in his sec ion, wi h de ailed indings discussed in he ollowing subsec ions.
4.4.1 Fac o s Tha Impac Quali y Assessmen and Visual A en ion.
Tempo al In o ma ion. Pa icipan s (13) poin ed ou ha he licke ing e ec is he mos bo he -
some a i ac in ou DPC playback scene, o en leading o lowe a ings. (P21: “I i ’s e y like ague
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:19
o like wi h big do s, hen i ’s ok, i ’s ake, bu i i licke s all he ime, ha could be a bi annoying,
i ’s so o a lo o see.”) Addi ionally, pa icipan s (20) epo ed ha hey ended o explo e mo e
du ing he obse a ion o high-mo ion poin cloud sequences.
Geome y and Tex u e. Dis o ions in geome y (12) and ex u e (11) a e iden i ied as he second
and hi d ac o s in luencing he subjec i e a ing o poin clouds unde sc u iny. (P3: “I was
obse ing p ecisely wo hings, he edges o he body and how dis o ed hey a e, and also some
dis o ions inside he cos ume.”) Fo DPC quali y assessmen , TI eme ges as mo e c i ical han ei he
geome ic o ex u e dis o ions, wi h geome y and ex u e exhibi ing nea ly equal impo ance.
Dis ance Impac s he Quali y Ra ing. Pa icipan s (8) disco e ed ha he iewing dis ance can
impac he subjec i e a ing o he same con en . Fi e o hem no ed ha he appea ance o holes is
de e mined by he iewing dis ance, esul ing in isually dis inc poin clouds e en when inspec ing
he same sequence. This inding complemen s he conclusion in [56] ha objec s iewed om a
g ea e dis ance end o ecei e lowe a ings compa ed o hei close coun e pa s.
Rela ionship be ween Visual A en ion and Quali y Assessmen . Pa icipan s a o ed he longd ess
(15), soldie (11), and dance (9) poin cloud sequences among all he con en s. The wo p ima y
easons ci ed we e hei high quali y (19) and he p esence o cues aiding in quali y sco e de e mi-
na ion (15). The cha ac e is ics o he poin cloud i sel in luenced pa icipan s’ isual a en ion.
Fo ins ance, indi iduals ended o ocus mo e on con en cha ac e ized by high mo ion (dance ),
ealism (longd ess,soldie ), and in ica e de ails (soldie ) o acili a e he quali y assessmen ask.
4.4.2 Subjec i e Quali y Expe imen Design o Da ase Cons uc ion.
P ocedu e. Pa icipan s (9) ecommend s eamlining he calib a ion and e o p o iling p ocess o
enhance use - iendliness, while also acknowledging he impo ance o main aining da a accu acy.
This sugges s a need o s iking a balance be ween da a p ecision and ease o use. Such a balanced
app oach is c ucial o ad ancing he de elopmen o eye- acking echniques, pa icula ly in e ms
o calib a ion p ocedu es.
In e ac ion wi h Con en . Pa icipan s (5) highligh ed he impo ance o allowing indi iduals
o de e mine he numbe o loops o each con en hemsel es. They emphasized ha o asks
equi ing quali y assessmen , hey may al eady ha e ins an esul s in mind o ce ain con en .
Fu he mo e, pa icipan s exp essed a desi e o inc eased in e ac ion be ween hemsel es and
he objec s wi hin DPCs playback scenes in VR o be e e alua ion o he quali y, unde sco ing
he impo ance o cus omizable in e ac ion o e ec i e e alua ion. (P4: “… I like o o a e i o
wha e e angle I wan and hen go and see i .”)
4.4.3 Con en Cha ac e is ics and Quali y Assessmen Task: In luencing Use In e ac ion.
Impac s o he Quali y Assessmen Task o In e ac ion in VR. Pa icipan s (21) a ibu ed mos
o hei mo emen o he need o obse e he on ace o de e mine o echeck hei quali y
sco e a ings. Addi ionally, VR cues p esen ed in he human igu es also p omp ed hem o mo e
ex ensi ely, enabling hem o iden i y mo e cues o e alua ing poin cloud quali y. (P5: “When I
was seeing he quali y, I was seeing he helme and i had like a small hing on op, and he e is a
di e ence in he quali y o ha as well, and e en he gun, you could see like di e en ea u es on
he gun. So he e a e mo e hings o look a .”)
Impac s o Con en s Cha ac e is ics o In e ac ion in VR. Pa icipan s (21) exp essed he iew
ha i he quali y le el is easy o disce n, hey p e e o emain s a iona y un il he sequence
comple es i s loop, such as in cases o excellen and poo quali y scales. Howe e , i he quali y
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:20 X. Zhou e al.
le el is ela i ely challenging o dis inguish, indi iduals op o mo e a ound o a comp ehensi e
obse a ion, e en wi hou engaging in andom o a ion ope a ions. This obse a ion aligns wi h
Damme’s conclusion [56] ha human pe cep ion o unde lying quali y ep esen a ions is in ica ely
linked wi h he con en and i s geome ic p ope ies unde examina ion.
5 Discussion
5.1 Ad ancing and Enhancing he Explana ion o DPC Quali y Assessmen Me ics
The e is po en ial o imp o ing he pe o mance o poin -based me ics h ough a s a egic
combina ion o isual a en ion and pooling me hods. Howe e , he isual a en ion da a a ailable
o DPCs ypically co e only a small egion o he dense objec , limi ing i s e icacy in ad ancing
DPC quali y assessmen me ics. Ou in es iga ion e eals ha employing empo al a ia ion
pooling me hods leads o dec eased pe o mance, p omp ing u he explo a ion in o sui able
empo al pooling echniques o spa ial- empo al pooling o DPC quali y me ics. I is c ucial o
accoun o he in insic cha ac e is ics o DPCs beyond me ely elying on ideo quali y indica o s.
Addi ionally, ca e ul conside a ion should be gi en o ma ching di e en dis o ion ypes wi h
app op ia e empo al pooling me hods.
5.2 The In luence o Task o DPC Visual A en ion
Ou expe imen , which ocuses on e alua ing he isual quali y o DPCs, likely in luenced pa -
icipan s’ a en ion owa d speci ic con en a eas ha acili a ed he ask: o example, a eas wi h
pa e ns on which dis o ions would easily be spo ed. Tha does no necessa ily mean ha he same
a ea would be a salien egion had he es been adminis e ed wi h a di e en ask o ask- ee.
Insigh s om he semi-s uc u ed in e iews con i m ha he iden i ied salien egions a e no only
inhe en ly a en ion-g abbing bu also o e cues ha aid pa icipan s in he quali y assessmen
p ocess. This dual in luence highligh s he con ex ual na u e o isual a en ion, d i en by bo h
in insic con en cha ac e is ics and he ask’s demands. Mo eo e , he accu acy o isual a en ion
map p edic ions a ies depending on he display en i onmen and associa ed asks [29,66]. Fu u e
esea ch should explo e how isual a en ion a ies ac oss di e en asks o ask- ee scena ios
o unco e mo e gene alized pa e ns o saliency in DPCs and hei implica ions o eal-wo ld
applica ions [10].
5.3 Visual A en ion Applica ions o DPC
The insigh s de i ed om gaze da a o DPC ex end a beyond quali y assessmen . Accu a e p e-
dic ion o isual a en ion p o ides a obus ounda ion o op imizing comp ession and s eaming
s a egies. Fo example, saliency-d i en bi alloca ion can enhance encoding e iciency by p io i iz-
ing he ideli y o egions ha cap u e use a en ion while alloca ing ewe esou ces o non-salien
a eas. Addi ionally, in eg a ing saliency maps wi h empo al mo ion in o ma ion, such as mo ion
ec o s, enables adap i e s eaming pa ame e s ha dynamically adjus o use ocus, ensu ing a
seamless and engaging expe ience. Fu he mo e, le e aging hese insigh s can suppo seman ic
segmen a ion o mo ion-dominan and mo ion-s a ic egions, which has signi ican po en ial o
applica ions such as objec ecogni ion and in e ac i e XR en i onmen s.
5.4 Da ase Applica ions and P ospec i e Ex ensions
QAVA-DPC, encompassing MOS/DMOS, use s’ gaze da a, and ou me iculously p ocessed isual
a en ion maps, hold signi ican po en ial as a ounda ional e e ence o he mul iple aspec s. Fi s ,
since he da ase includes he aw da a alongside he isual a en ion maps, i p o ides esea che s
and p ac i ione s wi h aluable esou ces o de elop and es no el algo i hms o pos -p ocessing
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:21
gaze da a and c ea ing isual a en ion maps. Addi ionally, QAVA-DPC acili a es he compa ison
o isual saliency maps ac oss di e en de ices (e.g., sc een-based o XR-based), highligh ing he
need o simila i y me ics ailo ed o DPC. Fu he mo e, he da ase enables he de elopmen
o objec i e quali y me ics and isual a en ion p edic ion models o DPC wi hou equi ing
esou ce-in ensi e use s udies. Mo eo e , insigh s de i ed om he quali a i e analysis and isual
a en ion design pa adigms can d i e ad ancemen s in DPC- ela ed esea ch and applica ions.
Finally, exis ing poin -based objec i e quali y me ics can be e ined and ailo ed o DPC o explo e
how o inco po a e isual a en ion and assess i s added alue in DPC quali y assessmen .
6 Conclusion
This a icle has p esen ed a da ase con aining subjec i e opinion sco es and isual a en ion maps
o DPCs in 6 DoF. An expe imen al p o ocol was designed o he subjec i e quali y assessmen and
isual a en ion de ec ion o he DPCs in a VR en i onmen . Addi ionally, we e alua ed exis ing
objec i e quali y me ics wi h and wi hou he in eg a ion o isual a en ion maps o s udy he
added alue o isual a en ion in objec i e DPC quali y me ics. Resul s om semi-s uc u ed
in e iews and he cap u ed isual a en ion maps p o ided deepe insigh s in o human beha io
in VR en i onmen s. While p e ious li e a u e indica es ha isual a en ion may be help ul o
image/ ideo s eaming, comp ession, as well as quali y assessmen , e alua ions and benchma king
o DPCs in ou wo k a e jus he beginning. In ou u u e wo k, we will ocus on p edic ing he
isual a en ion o DPCs and designing objec i e me ics o u he imp o e he me ics’ co ela ions
be ween p edic ed quali y sco es and g ound- u h subjec i e sco es.
Re e ences
[1]
Ali Ak, Emin Ze man, Suiyi Ling, Pa ick Le Calle , and Aljosa Smolic. 2021. The e ec o empo al sub-sampling
on he accu acy o olume ic ideo quali y assessmen . In 2021 Pic u e Coding Symposium (PCS), 1–5.
DOI:
h ps:
//doi.o g/10.1109/PCS50896.2021.9477449
[2]
E angelos Alexiou and Tou adj Eb ahimi. 2020. Towa ds a poin cloud s uc u al simila i y me ic. In 2020 IEEE
In e na ional Con e ence on Mul imedia Expo Wo kshops (ICMEW), 1–6.
DOI:
h ps://doi.o g/10.1109/ICMEW46912.
2020.9106005
[3]
E angelos Alexiou, Yana Nehmé, Emin Ze man, I ene Viola, Guillaume La oué, Ali Ak, Aljosa Smolic, Pa ick Le
Calle , and Pablo Cesa . 2023. Subjec i e and objec i e quali y assessmen o olume ic ideo. In Imme si e Video
Technologies. Else ie , 501–552.
[4]
E angelos Alexiou, I ene Viola, Tomás M. Bo ges, Tiago A. Fonseca, Rica do L. de Quei oz, and Tou adj Eb ahimi. 2019.
A comp ehensi e s udy o he a e-dis o ion pe o mance in MPEG poin cloud comp ession. APSIPA T ansac ions on
Signal and In o ma ion P ocessing 8, 1 (2019), e27. DOI: h ps://doi.o g/10.1017/ATSIP.2019.20
[5]
E angelos Alexiou, Peisen Xu, and Tou adj Eb ahimi. 2019. Towa ds modelling o isual saliency in poin clouds o
imme si e applica ions. In 2019 IEEE In e na ional Con e ence on Image P ocessing (ICIP). IEEE, 4325–4329.
[6]
Jochen An kowiak, T. D. F. Jamal Baina, F ance Vi o io Ba oncini, Noel Cha eau, F ance F anceTelecom, An onio Clau-
dio F ança Pessoa, F. U. B. S ephanie Colonnese, I aly Lau a Con in, Jo ge Ca iedes, and F ance Philips. 2000. Final e-
po om he ideo quali y expe s g oup on he alida ion o objec i e models o ideo quali y assessmen ma ch 2000.
Re ie ed om h ps://www.academia.edu/2102517/FINAL_REPORT_FROM_THE_VIDEO_QUALITY_EXPERTS_
GROUP_ON_THE_VALIDATION_OF_OBJECTIVE_MODELS_OF_VIDEO_QUALITY_ASSESSMENT_Ma ch_
[7]
Isayas B. Adhanom, Saman ha C. Lee, Eelke Folme , and Paul MacNeilage. 2020. Gazeme ics: An open-sou ce ool o
measu ing he da a quali y o HMD-based eye acke s. In ACM Symposium on Eye T acking Resea ch and Applica ions,
1–5.
[8]
Keming Cao, Yi Xu, and Pamela Cosman. 2020. Visual quali y o comp essed mesh and poin cloud sequences. IEEE
Access 8 (2020), 171203–171217.
[9]
Qiangqiang Cheng, Pengyu Sun, Chunsheng Yang, Yubin Yang, and Pe e Xiaoping Liu. 2020. A mo phing-based 3D
poin cloud econs uc ion amewo k o medical image p ocessing. Compu e Me hods and P og ams in Biomedicine
193 (2020), 105495.
[10]
Shiwei Cheng, Jing Fan, and Yilin Hu. 2023. Visual saliency model based on c owdsou cing eye acking da a and i s
applica ion in isual design. Pe sonal and Ubiqui ous Compu ing 27, 3 (2023), 613–630.
[11] J. H. Cla k. 1924. The Ishiha a es o colo blindness. Ame ican Jou nal o Physiological Op ics 5 (1924), 269–276.
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:22 X. Zhou e al.
[12]
Samuel Rhys Cox, May Lim, and Wei Tsang Ooi. 2023. VOLVQAD: An MPEG V-PCC olume ic ideo quali y
assessmen da ase . In P oceedings o he 14 h Con e ence on ACM Mul imedia Sys ems, 357–362.
[13]
E wan J. Da id, Jesús Gu ié ez, An oine Cou o , Ma hieu Pe ei a Da Sil a, and Pa ick Le Calle . 2018. A da ase o
head and eye mo emen s o 360 ideos. In P oceedings o he 9 h ACM Mul imedia Sys ems Con e ence, 432–437.
[14]
Eugene d’Eon, Bob Ha ison, Taos Mye s, and Philip A. Chou. 2019. JPEG Pleno Da abase: 8i Voxelized Full Bodies
(8iVFB 2)—A Dynamic Voxelized Poin Cloud Da ase .
[15]
Ma in Es e , Hans-Pe e K iegel, Jö g Sande , and Xiaowei Xu. 1996. A densi y-based algo i hm o disco e ing
clus e s in la ge spa ial da abases wi h noise. In P oceedings o he 2nd In e na ional Con e ence on Knowledge Disco e y
and Da a Mining, 226–231.
[16]
Yu Fan, Zicheng Zhang, Wei Sun, Xiongkuo Min, Jiaman Lin, Guang ao Zhai, and Ning Liu. 2023. MV-VVQA: Mul i-
iew lea ning o no- e e ence olume ic ideo quali y assessmen . In 2023 31s Eu opean Signal P ocessing Con e ence
(EUSIPCO). IEEE, 670–674.
[17]
F ede ick L. Fe is, III, Aa on Kasso , Geo ge H. B esnick, and Ian Bailey. 1982. New isual acui y cha s o clinical
esea ch. Ame ican Jou nal o Oph halmology 94, 1 (1982), 91–96.
[18]
Ped o Ga cia F ei as, Ma eus Gonçal es, Johann Homonnai, Ra ael Diniz, Mylène, and C. Q. Fa ias. 2022. On he
pe o mance o empo al pooling me hods o quali y assessmen o dynamic poin clouds. In 2022 14 h In e na ional
Con e ence on Quali y o Mul imedia Expe ience (QoMEX). IEEE, 1–6.
[19]
Ped o G. F ei as, Gio ani D. Luca o, Ma eus Gonçal es, Johann Homonnai, Ra ael Diniz, Mylène, and C. Q. Fa ias. 2022.
Compa a i e e alua ion o empo al pooling me hods o no- e e ence quali y assessmen o dynamic poin clouds. In
P oceedings o he 1s Wo kshop on Pho o ealis ic Image and En i onmen Syn hesis o Mul imedia Expe imen s, 35–41.
[20] Jesús Gu ié ez, Gulzhana Dandyye a, Ma eo Dal Mag o, Ca los Co és, Michele B izzi, Ma co Ca li, and Fede ica
Ba is i. 2023. Subjec i e e alua ion o dynamic poin clouds: Impac o comp ession and explo a ion beha io . In
2023 31s Eu opean Signal P ocessing Con e ence (EUSIPCO). IEEE, 675–679.
[21]
Jesus Gu ie ez, Pablo Pe ez, Ma a O duna, Ashu osh Singla, Ca los Co es, P ami Mazumda , I ene Viola, Kjell
B unns om, Fede ica Ba is i, Na alia Cieplinska, e al. 2021. Subjec i e e alua ion o isual quali y and simula o
sickness o sho 360◦ ideos: ITU-T ec. P. 919. IEEE T ansac ions on Mul imedia 24 (2021), 3087–3100.
[22]
Lau en I i, Ch is o Koch, and E ns Niebu . 1998. A model o saliency-based isual a en ion o apid scene analysis.
IEEE T ansac ions on Pa e n Analysis and Machine In elligence 20, 11 (1998), 1254–1259.
[23]
ITU-R BT.500-13. 2012. Me hodology o he Subjec i e Assessmen o he Quali y o Tele ision Pic u es. In e na ional
Telecommunica ions Union.
[24] ITU-T P.910. 2008. Subjec i e Video Quali y Assessmen Me hods o Mul imedia Applica ions. In e na ional Telecom-
munica ion Union.
[25]
ITU-T P.913. 2016. Me hods o he Subjec i e Assessmen o Video Quali y, Audio Quali y and Audio isual Quali y o
In e ne Video and Dis ibu ion Quali y Tele ision in Any En i onmen . In e na ional Telecommunica ion Union.
[26]
Ali eza Ja ahe i, Ca a ina B i es, Fe nando Pe ei a, and João Ascenso. 2021. Poin cloud ende ing a e coding:
Impac s on subjec i e and objec i e quali y. IEEE T ansac ions on Mul imedia 23 (2021), 4049–4064.
DOI:
h ps:
//doi.o g/10.1109/TMM.2020.3037481
[27]
Yize Jin, Meixu Chen, Todd Goodall, Anjul Pa ney, and Alan C. Bo ik. 2021. Subjec i e and objec i e quali y assessmen
o 2D and 3D o ea ed ideo comp ession in i ual eali y. IEEE T ansac ions on Image P ocessing 30 (2021), 5905–5919.
[28]
Robe S. Kennedy, No man E. Lane, Ke in S. Be baum, and Michael G. Lilien hal. 1993. Simula o sickness ques ion-
nai e: An enhanced me hod o quan i ying simula o sickness. The In e na ional Jou nal o A ia ion Psychology 3, 3
(1993), 203–220.
[29]
Haksub Kim, Sanghoon Lee, and Alan Con ad Bo ik. 2014. Saliency p edic ion on s e eoscopic ideos. IEEE T ansac ions
on Image P ocessing 23, 4 (2014), 1476–1490. DOI: h ps://doi.o g/10.1109/TIP.2014.2303640
[30]
Se gi Fe nández Langa, Ma io Mon agud, Gianluca Ce niglia o, and Da id Rincón Ri e a. 2022. Mul ipa y holomee -
ings: Towa d a new e a o low-cos olume ic holog aphic mee ings in i ual eali y. IEEE Access 10 (2022),
81856–81876.
[31]
Oli ie Le Meu , Alexand e Ninassi, Pa ick Le Calle , and Dominique Ba ba. 2010. O e isual a en ion o ee-
iewing and quali y assessmen asks: Impac o he egions o in e es on a ideo quali y me ic. Signal P ocessing:
Image Communica ion 25, 7 (2010), 547–558.
[32]
Zhi Li, Anne Aa on, Ioannis Ka sa ounidis, Anush Moo hy, and Megha Manoha a. 2016. Towa d a P ac ical Pe cep ual
Video Quali y Me ic. Re ie ed om h ps://ne lix echblog.com/ owa d-a-p ac ical-pe cep ual- ideo-quali y-me ic-
653 208b9652
[33]
Weisi Lin and C.-C. Jay Kuo. 2011. Pe cep ual isual quali y me ics: A su ey. Jou nal o Visual Communica ion and
Image Rep esen a ion 22, 4 (2011), 297–312.
[34]
Han ao Liu and Ing id Heynde ickx. 2011. Visual a en ion in objec i e image quali y assessmen : Based on eye-
acking da a. IEEE T ansac ions on Ci cui s and Sys ems o Video Technology 21, 7 (2011), 971–982.
DOI:
h ps:
//doi.o g/10.1109/TCSVT.2011.2133770
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
Subjec i e and Objec i e Quali y Assessmen o DPC 230:23
[35]
Zhongkang Lu, Weisi Lin, Xiaokang Yang, EePing Ong, and Susu Yao. 2005. Modeling isual a en ion’s modula o y
a e e ec s on isual sensi i i y and quali y e alua ion. IEEE T ansac ions on Image P ocessing 14, 11 (2005), 1928–1942.
[36]
Moi a Magui e and B id Delahun . 2017. Doing a hema ic analysis: A p ac ical, s ep-by-s ep guide o lea ning and
eaching schola s. All I eland Jou nal o Highe Educa ion 9, 3 (2017).
[37]
Jean-Eudes Ma ie, Yana Nehmé, Danillo G aziosi, and Guillaume La oué. 2023. C a ing he MPEG me ics o
objec i e and pe cep ual quali y assessmen o olume ic ideos. Quali y and Use Expe ience 8, 1 (2023), 4.
[38]
Ru ael Meku ia, Kees Blom, and Pablo Cesa . 2017. Design, implemen a ion, and e alua ion o a poin cloud codec
o ele-imme si e ideo. IEEE T ansac ions on Ci cui s and Sys ems o Video Technology 27, 4 (2017), 828–842.
DOI:
h ps://doi.o g/10.1109/TCSVT.2016.2543039
[39]
Gab iel Meyne , Yana Nehmé, Julie Digne, and Guillaume La oué. 2020. PCQM: A ull- e e ence quali y me ic o
colo ed 3D poin clouds. In 2020 12 h In e na ional Con e ence on Quali y o Mul imedia Expe ience (QoMEX). IEEE,
1–6.
[40]
Jesús Mo eno-A jonilla, Al onso López-Ruiz, J. Robe o Jiménez-Pé ez, José E. Callejas-Aguile a, and Juan M. Ju ado.
2024. Eye- acking on i ual eali y: A su ey. Vi ual Reali y 28, 1 (2024), 38.
[41]
Anh Nguyen and Zhisheng Yan. 2019. A saliency da ase o 360-deg ee ideos. In P oceedings o he 10 h ACM
Mul imedia Sys ems Con e ence, 279–284.
[42]
Minh Nguyen, Shi i Va s, Xuemei Zhou, I ene Viola, Pablo Cesa , Ch is ian Timme e , and He mann Hellwagne .
2024. ComPEQ-MR: Comp essed poin cloud da ase wi h eye acking and quali y assessmen in mixed eali y.
In P oceedings o he 15 h ACM Mul imedia Sys ems Con e ence (MMSys ’24). ACM, New Yo k, NY, 367–373.
DOI:
h ps://doi.o g/10.1145/3625468.3652182
[43]
Alexand e Ninassi, Oli ie Le Meu , Pa ick Le Calle , and Dominique Ba ba. 2009. Conside ing empo al a ia ions
o spa ial isual dis o ions in ideo quali y assessmen . IEEE Jou nal o Selec ed Topics in Signal P ocessing 3, 2 (2009),
253–265.
[44]
Cag i Ozcina and Aljosa Smolic. 2018. Visual a en ion in omnidi ec ional ideo o i ual eali y applica ions. In
2018 10 h In e na ional Con e ence on Quali y o Mul imedia Expe ience (QoMEX). IEEE, 1–6.
[45]
Yashas Rai, Jesús Gu ié ez, and Pa ick Le Calle . 2017. A da ase o head and eye mo emen s o 360 deg ee images.
In P oceedings o he 8 h ACM on Mul imedia Sys ems Con e ence, 205–210.
[46] Ignacio Reima , E angelos Alexiou, Jack Jansen, I ene Viola, Shishi Sub amanyam, and Pablo Cesa . 2021. CWIPC-
SXR: Poin cloud dynamic human da ase o social XR. In P oceedings o he 12 h ACM Mul imedia Sys ems Con e ence,
300–306.
[47]
Da io D. Sal ucci and Joseph H. Goldbe g. 2000. Iden i ying ixa ions and saccades in eye- acking p o ocols. In
P oceedings o he 2000 Symposium on Eye T acking Resea ch and Applica ions, 71–78.
[48]
E ich Schube , Jö g Sande , Ma in Es e , Hans Pe e K iegel, and Xiaowei Xu. 2017. DBSCAN e isi ed, e isi ed:
Why and how you should (s ill) use DBSCAN. ACM T ansac ions on Da abase Sys ems 42, 3 (2017), 1–21.
[49]
Sebas ian Schwa z, Ma ius P eda, Vi o io Ba oncini, Madhuka Budaga i, Pablo Cesa , Philip A. Chou, Robe
A. Cohen, Maja K i okuća, Sébas ien Lasse e, Zhu Li, e al. 2019. Eme ging MPEG s anda ds o poin cloud
comp ession. IEEE Jou nal on Eme ging and Selec ed Topics in Ci cui s and Sys ems 9, 1 (2019), 133–148.
DOI:
h ps:
//doi.o g/10.1109/JETCAS.2018.2885981
[50]
Ludwig Sidenma k, Ma hias N. Lys bæk, and Hans Gelle sen. 2023. GE-simula o : An open-sou ce ool o simula ing
eal- ime e o s o HMD-based eye acke s. In P oceedings o he 2023 Symposium on Eye T acking Resea ch and
Applica ions (ETRA ’23). ACM, New Yo k, NY, A icle 8, 6 pages. DOI: h ps://doi.o g/10.1145/3588015.3588417
[51]
Vincen Si zmann, Ana Se ano, Amy Pa el, Maneesh Ag awala, Diego Gu ie ez, Belen Masia, and Go don We zs ein.
2018. Saliency in VR: How do people explo e i ual en i onmen s? IEEE T ansac ions on Visualiza ion and Compu e
G aphics 24, 4 (2018), 1633–1642. DOI: h ps://doi.o g/10.1109/TVCG.2018.2793599
[52] S. Schwa z. 2019. Common es condi ions o PCC. ISO/IEC JTC1/SC29/WG11 Doc. N18665.
[53]
Shishi Sub amanyam, Jie Li, I ene Viola, and Pablo Cesa . 2020. Compa ing he quali y o highly ealis ic digi al
humans in 3do and 6do : A olume ic ideo case s udy. In 2020 IEEE Con e ence on Vi ual Reali y and 3D Use
In e aces (VR). IEEE, 127–136.
[54]
Shishi Sub amanyam, I ene Viola, Jack Jansen, E angelos Alexiou, Alan Hanjalic, and Pablo Cesa . 2022. Subjec i e
QoE e alua ion o use -cen e ed adap i e s eaming o dynamic poin clouds. In 2022 14 h In e na ional Con e ence on
Quali y o Mul imedia Expe ience (QoMEX). IEEE, 1–6.
[55]
E ic M. To lig, E angelos Alexiou, Tiago A. Fonseca, Rica do L. de Quei oz, and Tou adj Eb ahimi. 2018. A no el
me hodology o quali y assessmen o oxelized poin clouds. In Applica ions o Digi al Image P ocessing XLI. SPIE,
174–190.
[56]
Sam Van Damme, Imen Mahdi, Heman h Kuma Ra u i, Je oen an de Hoo , Filip De Tu ck, and Ma ia To es
Vega. 2023. Imme si e and in e ac i e subjec i e quali y assessmen o dynamic olume ic meshes. In 2023 15 h
In e na ional Con e ence on Quali y o Mul imedia Expe ience (QoMEX). IEEE, 141–146.
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
230:24 X. Zhou e al.
[57]
Sam Van Damme, Ma ia To es Vega, and Filip De Tu ck. 2021. A ull-and no- e e ence me ics accu acy analysis o
olume ic media s eaming. In 2021 13 h In e na ional Con e ence on Quali y o Mul imedia Expe ience (QoMEX).
IEEE, 225–230.
[58]
Je oen an de Hoo , Ma ia To es Vega, Ch is ian Timme e , Ali C. Begen, Filip De Tu ck, and Raimund Scha z. 2020.
Objec i e and subjec i e QoE e alua ion o adap i e poin cloud s eaming. In 2020 12 h In e na ional Con e ence on
Quali y o Mul imedia Expe ience (QoMEX), 1–6. DOI: h ps://doi.o g/10.1109/QoMEX48832.2020.9123081
[59]
Anouk an Kas e en, Kjell B unns öm, John Hedlund, and Ch is Snijde s. 2022. Quali y o expe ience o 360
ideo–subjec i e and eye- acking assessmen o encoding and eezing dis o ions. Mul imedia Tools and Applica ions
81, 7 (2022), 9771–9802.
[60]
I. Viola and P. S. Césa Ga cia. 2020. PCM_RR: A educed e e ence me ic o isual quali y e alua ion o poin cloud
con en s.
[61]
I ene Viola, Jack Jansen, Shishi Sub amanyam, Ignacio Reima , and Pablo Cesa . 2023. VR2Ga he : A collabo a i e
social VR sys em o adap i e mul i-pa y eal- ime communica ion. IEEE Mul iMedia 30, 2 (2023), 48–59. DOI:
10.1109/MMUL.2023.3263943
[62]
I ene Viola, Shishi Sub amanyam, and Pablo Cesa . 2020. A colo -based objec i e quali y me ic o poin cloud
con en s. In 2020 12 h In e na ional Con e ence on Quali y o Mul imedia Expe ience (QoMEX). IEEE, 1–6.
[63]
I ene Viola, Shishi Sub amanyam, Jie Li, and Pablo Cesa . 2022. On he impac o assessmen on he quali y o
expe ience o highly ealis ic digi al humans. Quali y and Use Expe ience 7, 1 (2022).
[64] Xinyi Wang, Angeliki Ka senou, and Da id Bull. 2023. UGC quali y assessmen : Explo ing he impac o saliency in
deep ea u e-based quali y assessmen . In Applica ions o Digi al Image P ocessing XLVI. SPIE, 351–365.
[65]
Heino Widdel. 1984. Ope a ional p oblems in analysing eye mo emen s. In Ad ances in Psychology. Vol. 22, Else ie ,
21–29.
[66]
Nan Wu, Kaiyan Liu, Ruizhi Cheng, Bo Han, and Puqi Zhou. 2024. Theia: Gaze-d i en and pe cep ion-awa e olume ic
con en deli e y o mixed eali y headse s. In P oceedings o he 22nd Annual In e na ional Con e ence on Mobile
Sys ems, Applica ions and Se ices, 70–84.
[67]
Xinju Wu, Yun Zhang, Chunling Fan, Junhui Hou, and Sam Kwong. 2021. Subjec i e quali y da abase and objec i e
s udy o comp essed poin clouds wi h 6DoF head-moun ed display. IEEE T ansac ions on Ci cui s and Sys ems o
Video Technology 31, 12 (2021), 4630–4644. DOI: h ps://doi.o g/10.1109/TCSVT.2021.3101484
[68]
Yi Xu, Yao Lu, and Ziyu Wen. 2017. Owlii dynamic human ex u ed mesh sequence da ase . In ISO/IEC JTC1/SC29/WG1
1 inpu documen m41658.
[69]
Mengyu Yang, Di Wu, Zelong Wang, Miao Hu, and Yipeng Zhou. 2023. Unde s anding and imp o ing pe cep ual
quali y o olume ic ideo s eaming. In 2023 IEEE In e na ional Con e ence on Mul imedia and Expo (ICME). IEEE,
1979–1984.
[70]
Emin Ze man, Pan Gao, Cag i Ozcina , and Aljosa Smolic. 2019. Subjec i e and objec i e quali y assessmen o
olume ic ideo comp ession. Elec onic Imaging 10 (2019), 323–321.
[71]
Emin Ze man, Cag i Ozcina , Pan Gao, and Aljosa Smolic. 2020. Tex u ed mesh s colou ed poin cloud: A subjec i e
s udy o olume ic ideo comp ession. In 2020 12 h In e na ional Con e ence on Quali y o Mul imedia Expe ience
(QoMEX), 1–6. DOI: h ps://doi.o g/10.1109/QoMEX48832.2020.9123137
[72]
Wei Zhang and Han ao Liu. 2017. S udy o saliency in objec i e ideo quali y assessmen . IEEE T ansac ions on Image
P ocessing 26, 3 (2017), 1275–1288. DOI: h ps://doi.o g/10.1109/TIP.2017.2651410
[73]
Wei Zhang and Han ao Liu. 2017. Towa d a eliable collec ion o eye- acking da a o image quali y esea ch:
Challenges, solu ions, and applica ions. IEEE T ansac ions on Image P ocessing 26, 5 (2017), 2424–2437.
[74]
Yihuan Zhang, Liang Wang, and Yi an Dai. 2023. PLOT: A 3D poin cloud objec de ec ion ne wo k o au onomous
d i ing. Robo ica 41 (2023), 1–17.
[75]
Xuemei Zhou, I ene Viola, E angelos Alexiou, Jack Jansen, and Pablo Cesa . 2023. QAVA-DPC: Eye- acking based
quali y assessmen and isual a en ion da ase o dynamic poin cloud in 6 DoF. In 2023 IEEE In e na ional Symposium
on Mixed and Augmen ed Reali y (ISMAR), 69–78. DOI: h ps://doi.o g/10.1109/ISMAR59233.2023.00021
Recei ed 4 Augus 2024; e ised 30 Janua y 2025; accep ed 11 Ma ch 2025
ACM T ans. Mul imedia Compu . Commun. Appl., Vol. 21, No. 8, A icle 230. Publica ion da e: Augus 2025.
