Statistical signal processing techniques for visual post-production
François Pitié · Arrow@dit (Dublin Institute of Technology) · 2007
Digital media post-production is an industry standard step in media creation.Now that issues of speed and physical storage have largely been rendered less problematic, the emphasis has shifted toward increasing levels of automation.This thesis makes several contributions in the domain of visual post-production in an attempt to bring advanced statistical digital signal processing techniques to bear on several key problem.The problem of transfer of statistics is rst addressed.It transpires that this fundamental process underlies several key problems in lm grading.This can be understood by considering an image as of features samples.Then texture or colour aspects can be described by the statistics of texture or colour samples.The rst and major problem is If you are reading this page, you are probably looking for your name, and you most likely deserve it, so thank you very much!Else I believe you are suering from compulsive insanity.I would like to thank all former and present members of the Sigmedia group who have helped me over the years, and especially my PhD companions: the exhilarating Hugh DenmanAnil bless his thesis one dayfor his huge and generous help, and Dr. Francis Kelly because he's a man from Carlow and for all the good time in and outside the lab.I wish to express my deep gratitude to my co-supervisor Dr. Rozenn Dahyot for the time and precious help she gave me.And of course, most of all, I would like to thank my supervisor Dr. Anil Kokaram for his patience and intelligence in deciphering my obscure drafts, the good atmosphere he instigates in the lab, his contagious motivation and, above all, his constant eorts and help over the last 3 years