Tim Bruylants
Total Page:16
File Type:pdf, Size:1020Kb
FACULTY OF ENGINEERING Departement of Electronics and Informatics Advanced Coding Technologies For Medical and Holographic Imaging Algorithms, Implementations and Standardization Thesis submitted in fulfilment of the requirements for the award of the degree of Doctor in de ingenieurswetenschappen (Doctor in Engineering) by Tim Bruylants July 2015 Advisor(s): Prof. Dr. Peter Schelkens Prof. Dr. Adrian Munteanu Examining Committee Prof. Dr. ir. Peter Schelkens { Vrije Universiteit Brussel { Promoter Prof. Dr. ir. Adrian Munteanu { Vrije Universiteit Brussel { Promoter Prof. Dr. ir. Leo Van Biesen { Vrije Universiteit Brussel { Committee chair Prof. Dr. ir. Johan Deconinck { Vrije Universiteit Brussel { Committee vice-chair Prof. Dr. Bart Jansen { Vrije Universiteit Brussel { Committee secretary Prof. Dr. ir. Søren Forchhammer { Technical University of Denmark { Member Prof. Dr. ir. Aleksandra Pizurica - Universiteit Gent - Member Prof. dr. Johan De Mey { Faculteit Geneeskunde, UZ Brussel { Member In memory of Bruno and Pam. \An expert is a man1 who has made all the mistakes that can be made in a very narrow field.” { Niels Bohr 1women are experts by default Table of contents Acknowledgmentsv Synopsis vii Acronyms ix 1 Introduction1 1.1 Motivation................................1 1.1.1 Volumetric medical image coding................1 1.1.2 Digital holographic image coding................4 1.1.3 JPEG Standardization......................6 1.2 Outline and major contributions....................7 2 Image coding overview9 2.1 Introduction................................9 2.2 Concepts and definitions......................... 10 2.2.1 Digital images.......................... 10 2.2.2 Entropy and mutual information................ 11 2.2.3 Quantization........................... 14 2.2.4 Lossless and lossy compression................. 19 2.2.5 Objective quality metrics.................... 22 2.2.6 Subjective quality metrics.................... 25 2.3 Discrete Wavelet Transform (DWT).................. 27 2.3.1 Classical or non-directional DWT................ 27 2.3.2 Directional DWT......................... 31 2.3.3 Intermediate conclusions..................... 32 2.4 JPEG................................... 32 2.4.1 Pre-processing.......................... 32 2.4.2 Discrete Cosine Transform (DCT)............... 34 2.4.3 Quantization........................... 35 i Table of contents 2.4.4 Entropy coding.......................... 36 2.5 JPEG-LS................................. 37 2.5.1 Context-based prediction.................... 38 2.5.2 Context modeling........................ 39 2.5.3 Golomb-Rice coding....................... 40 2.5.4 Arithmetic coding with Part 2................. 40 2.6 JPEG 2000................................ 42 2.6.1 Pre-processing.......................... 42 2.6.2 Discrete Wavelet Transform (DWT).............. 45 2.6.3 Quantization........................... 48 2.6.4 Tier-1: EBCOT.......................... 49 2.6.5 Code-stream organization.................... 53 2.6.6 Rate control............................ 57 2.6.7 Region of Interest (ROI) coding................. 60 2.6.8 Intermediate conclusions..................... 61 2.7 HEVC................................... 61 2.7.1 High-level syntax......................... 62 2.7.2 Intra-picture prediction..................... 64 2.7.3 Inter-picture prediction..................... 64 2.7.4 Transform and quantization................... 65 2.7.5 In-loop filters........................... 65 2.7.6 Entropy coding.......................... 66 2.8 Conclusions................................ 66 3 JP3D for volumetric image coding 67 3.1 Introduction................................ 67 3.2 Why JP3D?................................ 67 3.3 JP3D: going volumetric......................... 68 3.3.1 Preprocessing........................... 69 3.3.2 The 3D discrete wavelet transform (3D-DWT)........ 69 3.3.3 Quantization........................... 73 3.3.4 Bit-modeling and entropy coding................ 73 3.4 Bit-Stream Organization......................... 75 3.4.1 The three dimensional canvas coordinate system....... 75 3.4.2 Code-stream........................... 83 3.4.3 Rate Control........................... 87 3.5 Additional features of JP3D....................... 88 3.5.1 Region-of-Interest........................ 88 3.6 Implications for other parts of JPEG 2000............... 89 3.6.1 Volumetric extension to JPIP.................. 89 3.7 JP3D Verification Model Software................... 90 ii Table of contents 3.8 Volumetric context modeling (not part of JP3D)........... 90 3.8.1 Introduction........................... 90 3.8.2 3D context models........................ 91 3.8.3 Mutual information based context modeling.......... 92 3.8.4 Near-optimal context classification............... 94 3.8.5 Context model experiments................... 95 3.8.6 Context modeling conclusions.................. 99 3.9 Conclusions................................ 100 4 Volumetric image coding extensions 101 4.1 Introduction................................ 101 4.2 Investigated Extensions for JPEG 2000................. 104 4.2.1 Alternative wavelet filters.................... 104 4.2.2 Block-based volumetric Direction-Adaptive DWT (DA-DWT) 105 4.2.3 Generic segmentation based on SD-DA-DWT......... 107 4.2.4 Volumetric coding with block-based intra-band prediction.. 110 4.3 Volumetric compression framework, JP3D+DA............ 111 4.4 Experimental results........................... 114 4.4.1 Image data set.......................... 114 4.4.2 Volumetric decomposition structures.............. 114 4.4.3 Block-based Intra-band prediction............... 122 4.4.4 Enabling DA-DWT........................ 123 4.4.5 Comparison with H.265/MPEG-H HEVC and JPEG-LS... 128 4.4.6 Computational complexity.................... 132 4.4.7 Visual comparisons........................ 132 4.4.8 Summarizing rate-distortion performance tables........ 136 4.5 Conclusions................................ 142 5 Coding of holographic data 145 5.1 Introduction................................ 145 5.2 Representation of off-axis holograms.................. 145 5.3 Full Packet Decomposition with JPEG 2000.............. 149 5.4 Proposed extensions for JPEG 2000.................. 152 5.4.1 Truly Arbitrary Packet Decompositions............ 152 5.4.2 Directional Adaptive Discrete Wavelet Transform (DA-DWT) 155 5.5 Experiments................................ 157 5.5.1 Test Data............................. 157 5.5.2 Test setup and settings..................... 159 5.5.3 Results.............................. 160 5.6 Conclusions................................ 164 iii Table of contents 6 JPEG standardization 165 6.1 Introduction................................ 165 6.2 JPEG Committee............................. 165 6.3 JBIG, JPEG-LS and JPSearch..................... 167 6.4 JPEG 2000................................ 168 6.5 JPEG XR................................. 169 6.5.1 Introduction........................... 169 6.5.2 Technical overview........................ 169 6.5.3 Subjective testing methodology................. 171 6.6 JPEG Systems.............................. 178 6.6.1 Introduction........................... 178 6.6.2 JPEG Systems, a layered architecture............. 180 6.6.3 Conclusions on JPEG Systems................. 180 6.7 Advanced image coding and evaluations (AIC)............ 181 6.8 JPEG XT................................. 181 6.8.1 Introduction........................... 181 6.8.2 Suite of standards........................ 182 6.8.3 Forward compatibility and new functionality......... 183 6.8.4 JPEG XT Boxes......................... 184 6.8.5 High dynamic range coding................... 185 6.8.6 Lossless and near-lossless coding................ 187 6.8.7 Alpha channel coding...................... 188 6.8.8 Conclusions on JPEG XT.................... 189 6.9 Conclusions................................ 189 7 Epilogue 191 7.1 Conclusions................................ 191 7.2 Future research potential........................ 195 List of publications 199 References 205 Index 217 iv Acknowledgments Here I am, finally arriving at the end of a long but awarding period. Like any PhD, this work too is the result of years of labor, often in cooperation with various people. For this reason, I want to express my appreciation to everyone who contributed to this work in some way. However, before I continue with personal acknowledgments, I would like to state that my memory sometimes fails me on simple things, such as remembering names. Fact is, I would like to thank so many persons, but there is always the risk of forgetting someone. So, to mitigate this risk, I will keep these acknowledgments rather short and if you are that person that also deserved a credit, then don't feel offended. After obtaining my Master of Science degree, I started working for a relatively small, but nice company. However, after some years, I decided to make a career switch into the field of academic research. Yet, going from a company to the academic world is not evident, as I was to find out. Therefore, and foremost, I thank Peter for hiring me as a researcher and providing me with the opportunity and support to not only do a PhD, but also to become part of the wonderful world of standardization. I also genuinely thank Adrian as my co-advisor for his great help and valuable input to this research. And, with Peter and Adrian, I also thank the jury members for their participation and effort regarding this PhD. Of all my friends, I would specifically like to