New artificial life model for image enhancement

15 h In e na ional Con e ence on Expe imen al Mechanics ICEM15 1 PAPER REF: 3830 AN ARTIFICIAL LIFE MODEL FOR IMAGE ENHANCEMENT Alex F. de A aujo (*) , João Manuel R.S. Ta a es Ins i u o de Engenha ia Mecânica e Ges ão Indus ial, Faculdade de Engenha ia, Uni e sidade do Po o, Po ugal (*) Email: [email protected] ABSTRACT This pape p esen s an a i icial li e model o image enhancemen . The esul s o some adi ional enhancemen me hods we e analyzed and compa ed wi h he ones ob ained by he model p oposed. The quali a i e and quan i a i e es s pe o med allowed o conclude ha he new model is p omising, as is able o enhance ansi ions o he objec s p esen ed in he o iginal images and make hem mo e isually pe cep ible. INTRODUCTION The e a e se e al ac o s ha can con ibu e o damage he in o ma ion in images, such as loss o ocus, p esence o noise, e lec ions and shadows, and insu icien illumina ion. Image enhancemen me hods ha e been de eloped o educe he e ec o such damages, by imp o ing he con as be ween he objec s ep esen ed, emphasizing hei mo e signi ican ea u es (Hashemi, 2010). A i icial models, inspi ed on he biological p ocesses ha cha ac e ize li ing o ganisms, ha e been adop ed o pe o m compu a ional image analysis asks (Hama neh, 2009), (McIne ney, 2002). Such biological p ocesses include g owing, na u al selec ion, e olu ion, locomo ion and lea ning (Te zopoulos, 1999). This pape p esen s a new a i icial li e model, which is inspi ed on he beha io o an he bi o e o ganism when i is in an en i onmen and selec s i s ood, o image enhancemen . Thus, conside ing an en i onmen con aining he bs o di e en heigh s, he smalle he bs a e ea en i s , because hey a e smoo he and mo e nu i ional. The e o e, he e will be a endency o inc ease he di e ences be ween he sho e and alle he bs due o he mo ion and ea ing p ocess o he o ganism, in a simila way as i is desi ed in he image enhancemen . Quali a i e and quan i a i e compa isons pe o med on he esul s ob ained by he p oposed model and some image enhancemen adi ionally me hods allowed o conclude ha ou solu ion is p omising, being able o imp o e he quali y o he damaged images and hei isual pe cep ion conside ably. RESULTS AND CONCLUSIONS The PSNR (Peak Signal Noise Ra io) indices calcula ed om he compa ison o he o iginal images and ones ob ained by he p oposed model and some image enhancemen adi ionally me hods a e ep esen ed in Figu e 1. This igu e allows o ealize ha he adop ed model e u ned images wi h he bes indices. The image es se used was composed by syn ac ic images c ea ed using an image edi o and he well-known “Lena” and “Came amen” images, bo h o hem wi h he con as a ec ed by he addi ion o con olled noise and blu ing. Rega ding he “Lena image”, Figu e 2 depic s he enhancemen esul s, and Table 1 p esen s he associa ed PSNR indices. This wo k has shown ha he p oposed a i icial li e model o image enhancemen is p omising, leading o be e esul s han he enhancemen adi ionally me hods. To imp o e Po o/Po ugal, 22-27 July 2012 2 he e iciency o ou model, we in end o de elop an enhanced cogni i e sys em, and apply op imiza ion echniques and pa allel p og amming o speed up he compu a ional p ocess. Figu e 1 - G aph wi h he PSNR indices o he images es ed. Figu e 2 - Resul s o he enhancemen me hods applied o “Lena image”. Table 1- PSNR o “Lena” image. METHOD PSNR Equaliza ion o His og am 13.49 No maliza ion 18.36 Quad a ic Enhancemen 19.43 Squa e Roo Enhancemen 6.84 Loga i hmic Enhancemen 4.70 P oposed Me hod 20.99 ACKNOWLEDGMENTS This wo k was pa ially done in he scope o he p ojec s wi h e e ences PTDC/EEA- CRO/103320/2008, UTAus in/MAT/0009/2008 and UTAus in/CA/0047/2008, inancially suppo ed by FCT - Fundação pa a a Ciência e a Tecnologia in Po ugal. The i s au ho would like o hank his PhD g an om FCT wi h e e ence SFRH/BD/61983/2009. REFERENCES [1]-Hama neh G, McIn osh C, McIne ney T, Te zopoulos D. De o mable O ganisms: An A i icial Li e F amewo k o Au oma ed Medical Image Analysis. Book: Compu a ional In elligence In Medical Imaging: Techniques and Applica ions, 2009, p. 433-474. [2]-Hashemi S, Kiani S, No oozi N, Moghaddam ME. An image con as enhancemen me hod based on gene ic algo i hm. In Pa e n Recogni ion Le e s. 2010, Vol. 31:13, p. 1816-1824. [3]-McIne ney T, Hama neh G, Shen on M, Te zopoulos D. De o mable o ganisms o au oma ic medical image analysis. In Medical Image Analysis. 2002, Vol. 6:3, p. 251-266. [4]-Te zopoulos D. A i icial li e o compu e g aphics. In Communica ions o he ACM, 1999, Vol. 42:8, p. 32-42.