Presses universitaires de Louvain onixsuitesupport@onixsuite.com 20260717T2332Z eng

COM.ONIXSUITE.9782390615682 03 01 Presses universitaires de Louvain 01 SKU 108461 02 2390615686 03 9782390615682 15 9782390615682 02 01 00 BC 00 01 9.45 in 02 6.30 in 08 7.65 oz 01 24 cm 02 16 cm 08 217 gr BE 10 02 1025 01 02 Thèses de l'Université catholique de Louvain (UCL) 03 Thèses de l'École polytechnique de Louvain 01 01 Joint learned compression and denoising schemes for improved photographic image quality and reduced processing complexity 1 A01 01 Onixsuite Contributor ID 6923 Benoît Brummer Brummer, Benoît Benoît Brummer Benoit Brummer started his PhD thesis in 2020 at the University of Louvain's Institute for Information and Communication Technologies, Electronics and Applied Mathematics (ICTEAM), under the supervision of Professor Christophe De Vleeschouwer.. He holds a B.Sc. in Computer Engineering from the University of Central Florida and a M.Sc. in Computer Science from the University of Louvain. His research interests include image denoising and processing, optimization problems, and computational efficiency. He has contributed to the development of the Natural Image Noise Dataset. His extracurricular interests include outdoor adventures, photography, forest gardening, nutrition, video and board games, and open-source software. Benoit Brummer started his PhD thesis in 2020 at the University of Louvain's Institute for Information and Communication Technologies, Electronics and Applied Mathematics (ICTEAM), under the supervision of Professor Christophe De Vleeschouwer.. He holds a B.Sc. in Computer Engineering from the University of Central Florida and a M.Sc. in Computer Science from the University of Louvain. His research interests include image denoising and processing, optimization problems, and computational efficiency. He has contributed to the development of the Natural Image Noise Dataset. His extracurricular interests include outdoor adventures, photography, forest gardening, nutrition, video and board games, and open-source software. 01 eng 00 128 03 10 TEC000000 10 TEC015000 10 TEC043000 12 T 12 U 24 INTERNET Graphisme et multimédia 24 INTERNET Informatique 29 2012 3069 TECHNIQUES ET SCIENCES APPLIQUEES 29 2012 3072 Electronique 29 2012 3077 Graphisme et image 01 06 03 00 This thesis addresses image compression and denoising using deep learning, enhancing visual quality and computational efficiency. For compression, convolutional autoencoders are improved by replacing computationally expensive, single-feature parametric entropy models with a simplified scheme using multiple, pre-learned, static cumulative distribution function tables. This significantly reduces entropy coding/decoding complexity during inference. Image denoising involves the removal of unwanted noise from images captured under suboptimal conditions. Noise not only degrades image quality but also impairs the performance of compression algorithms, as it is inherently non-compressible. This thesis proposes a unified model that performs joint denoising and compression. By training the model on noisy-clean image pairs across a wide range of noise levels, it learns to denoise images as part of the compression process while maintaining the computational cost of compression alone. This joint approach improves rate-distortion performance compared to compressing noisy images or using separate denoising and compression models. Additionally, the model is capable of producing decompressed images with visual quality superior to that of the noisy uncompressed input. The final part of this thesis focuses on raw input images. Processing raw or minimally processed images offers substantial gains in both compression efficiency and denoising quality compared to working with fully processed images. Treating Bayer images as 4-channel inputs reduces the computational complexity of denoising models and compression encoders by a factor of four, while also improving compression performance at lower bitrates. Moreover, denoising raw or linear RGB images early in the processing pipeline enables greater generalization. A novel dataset of raw cleannoisy image pairs is introduced to support further research and the development of models integrated into image processing software pipelines. 02 00 This thesis addresses image compression and denoising using deep learning, enhancing visual quality and computational efficiency. 03 00 This thesis addresses image compression and denoising using deep learning, enhancing visual quality and computational efficiency. For compression, convolutional autoencoders are improved by replacing computationally expensive, single-feature parametric entropy models with a simplified scheme using multiple, pre-learned, static cumulative distribution function tables. This significantly reduces entropy coding/decoding complexity during inference. Image denoising involves the removal of unwanted noise from images captured under suboptimal conditions. Noise not only degrades image quality but also impairs the performance of compression algorithms, as it is inherently non-compressible. This thesis proposes a unified model that performs joint denoising and compression. By training the model on noisy-clean image pairs across a wide range of noise levels, it learns to denoise images as part of the compression process while maintaining the computational cost of compression alone. This joint approach improves rate-distortion performance compared to compressing noisy images or using separate denoising and compression models. Additionally, the model is capable of producing decompressed images with visual quality superior to that of the noisy uncompressed input. The final part of this thesis focuses on raw input images. Processing raw or minimally processed images offers substantial gains in both compression efficiency and denoising quality compared to working with fully processed images. Treating Bayer images as 4-channel inputs reduces the computational complexity of denoising models and compression encoders by a factor of four, while also improving compression performance at lower bitrates. Moreover, denoising raw or linear RGB images early in the processing pipeline enables greater generalization. A novel dataset of raw cleannoisy image pairs is introduced to support further research and the development of models integrated into image processing software pipelines. 02 00 This thesis addresses image compression and denoising using deep learning, enhancing visual quality and computational efficiency. 04 00 1 Introduction 1 1.1 Motivation 1 1.2 Thesis contributions 3 1.3 Thesis structure 5 2 Background 7 2.1 Image development 7 2.2 Convolutional neural networks 11 2.3 Image denoising 15 2.4 Image compression 22 3 End-to-end optimized image compression with competition of prior distributions 33 3.1 Related works 34 3.2 Competition of prior distributions 35 3.3 Experiments 39 3.4 Conclusion 45 4 On the importance of denoising when learning to compress images 47 4.1 Background 49 4.2 Jointly learned denoising and compression 51 4.3 Experiments 54 4.4 Conclusion 63 5 Learning joint denoising, debayering and compression from the Raw Natural Image Noise Dataset 65 5.1 Related work 67 5.2 Proposed approach 70 5.3 Validation methodology 77 5.4 Results and discussion 81 5.4.1 Denoising performance 81 5.5 Conclusion 89 6 Conclusion 91 6.1 Results summary 92 6.2 Perspectives 93 Acknowledgements 97 Bibliography 99 01 00 03 02 01 D502 02 2838 03 1892 06 0f0b719599d59d6510d528e95f5a21ca 07 479771 https://pul.uclouvain.be/resources/titles/29303100065230/images/b3ddb7c5b10be95dbc3f9152c58becce/HIGHQ/9782390615682.jpg 17 20260616T0733Z 02 01 D502 02 188 03 125 06 284c5c740622195241a1867327a491d4 07 24636 https://pul.uclouvain.be/resources/titles/29303100065230/images/b3ddb7c5b10be95dbc3f9152c58becce/THUMBNAIL/9782390615682.jpg 17 20260616T0733Z 06 3052405007518 Presses universitaires de Louvain 01 01 Dilicom PULOUVAIN 06 3052405007518 Presses universitaires de Louvain Louvain-la-Neuve BE 04 01 20260616 11 20260616 01 WORLD 02 01 GCOI 29303100065230 WORLD FR 04 03 06 3012405004818 CIACO - DUC 01 2 20 1 02 00 02 02 STD 02 18.00 01 S 21.00 14.88 3.12 EUR BE