Challenges In Video Quality Assessment For Underwater Wireless Sensor Networks

CHALLENGES IN VIDEO QUALITY ASSESSMENT FOR UNDERWATER WIRELESS SENSOR NETWORKS Javier Poncela J.M. Moreno-Roldán, M.A. Luque-Nieto, P. Otero Communications Engineering Department University of Málaga, Spain IWIEM’17 Motivations •Ocean scientists often need not only sensor measures (temperature, salinity…) but also they need to watch underwater environments. •Images from oceanic resources are currently difficult and expensive to obtain. – Exploration expeditions with divers or robots submerging with cameras are needed. 2 •Video services in USNs would allow to reduce these costs. IWIEM’17 Subjective Video Quality Assessment (VQA) • Subjective tests are considered the most reliable approach to quality. The opinion is gathered directly from users. It is a well known procedure (ITU BT.500 and P.910). •They also bring important drawbacks. It is a time-consuming method. oA session for a single user is at least 20 minutes. It requires a fair amount of human resources. oViewers should not be experienced in quality assessment. oIf the same people take part in different experiments, a wide enough time lapse must be used. 3IWIEM’17 Challenges in subjective VQA – Underwater • Underwater video is currently used for specific applications. – Ocean scientists, companies managing underwater resources, safety and security specialists… • It is more difficult to find an appropriate group of evaluators. • In this kind of application, quality is usually related to the tasks the video is used for. – Different professionals perform different tasks. The quality can be perceived in a different way. 4IWIEM’17 Challenges in objective VQA – Underwater 5 • Nodes are virtually unreachable once deployed. – The original unimpaired video cannot be recovered. • Nodes must operate as long as possible. Energy saving is a priority. – Intensive processing tasks that would reduce battery life should be avoided. • The low bitrate heavily constrains the amount of information that can be sent for quality measuring purposes. IWIEM’17 Objective Video Quality Assessment • The quality is estimated with a mathematical model. – Once the model is built and tested, quality can be assessed without the disadvantages of subjective VQA. – Most models compute an approximation to the MOS. • Models can be classified according to their inputs: –Full Reference –Reduced Reference. The received signal and some features from the original signal are used to compute the quality estimation. –No Reference. Only the received signal required for the quality estimation. 6           2 2 2 )ln()ln( exp1 Fr Ofr D OfrFr IMOS BrvvOfr 21  5 4 3 3 1 v Ofr v Br v vI           IWIEM’17 Applicability of VQA methods – Full Reference 7 • The received and the original signal are analyzed and compared to compute the quality estimation. • Extensively used • Standard Algorithms – PEVQ algorithm models human visual system (ITU J.247) – SSIM performs better than PEVQ (more recent) • Drawbacks – Original signal is required (UWSNs bitrates are too low) – Involve heavy image processing, expensive energy use for underwater nodes Usefulness only in laboratory tests IWIEM’17 Applicability of VQA methods – Reduced Reference 8 • The received signal and some features from the original signal are used to compute the quality estimation. • RR methods in J.249 use 15-256 kbps – Data for features should only need a fraction of video bitrate • Still too large bitrate • Feature extraction still requires intensive image processing – Energy concerns • New method uses 0.875 kbps (~6% of 15 kbps video flow) Applicable but requires novel algorithms IWIEM’17 Applicability of VQA methods – No Reference 9 • Advantages: – Only needs received signal • Pixel-based, bit-stream or network parameter analysis – No extra processing in intermediate nodes • Disadvantage: Good performing methods, but none tested in underwater channels • Standard parametric method G.1070 was intended for videoconferences – Cannot be extended to underwater (previous published study) due to mismatch in quality scores [10] Proposal: New parametric model for underwater IWIEM’17