Entropy Coding for Image Compression Based on Generalized Lifting and SPECK

Abstract

Image source coding became very important in the 80's. There were some key factors that helped to that data (image) compression revolution. First, the idea of digital image: an image composed by a set of niteprecision coe cients placed over a nite and discrete grid. To represent the image one could scan its coe cients and concatenate the coe cients binary representation. This strategy does not take into account any of the characteristics of the input signal (the image). Moreover it is not an e cient representation of the image and compression techniques are therefore necessary. Second, researchers had started to develop new techniques and softwares in the eld of text compression and generic data coding some years ago (see [10, 32]). Third, the presence of digital image processing in the market increased due to its simplicity, performance and results. Those are, under my opinion, the key factors that promoted image entropy coding at the research level in the 80's. On the other hand, this data (image) compression revolution helped to trigger another spectacular growth: the number of plain Internet users. A large amount of people started to share information: plain and formatted text, photo, music, video,... These information exchanges required better-performed techniques in terms of ratedistortion characteristics. User's needs increased and researchers, developers and producers had to upgrade the system's quality and capacity. This Internet growth, in the 1990s, would not have been possible without the previous image compression systems quality rise.