Digital Image Processing

E-book link: 数字图像处理（中）第三版 (1).pdf
Study Chapters 1, 2, 4, 6.1-6.2, 8; Chapter 8 can be combined with JPEG Encoding Details Introduction to understand the general encoding process.

Deep Learning

First, you need a sufficient understanding of Python. Optional e-book: Python编程：从入门到实践.pdf
Learn PyTorch, related course on Bilibili: Li Mu's AI Learning Space - Dive into Deep Learning PyTorch Edition - Bilibili Video (bilibili.com), focus on 00~29.2, 31, 33-37, 47, 47.2

Introductory Fundamental Papers

Combine papers and code (CompressAI) to try training a set of models yourself and plot the RD curve.

Common training and validation sets: ImageNet/COCO
Common test set: 24 Kodak images; due to edge anomalies in the original images, sometimes cropped square images are used.

1. Ballé, J., et al. (2015). "Density modeling of images using a generalized normalization transformation." arXiv preprint arXiv:1511.06281. The commonly used activation layer GDN in AI Codec, related code: CompressAI/compressai/layers/gdn.py at master · InterDigitalInc/CompressAI (github.com)
2. Ballé, J., et al. (2016). "End-to-end optimized image compression." arXiv preprint arXiv:1611.01704. Introduces the basic architecture of AI Codec, which can be viewed in conjunction with JPEG encoding, comparing the transform, quantization, entropy coding present in both processes, and understanding the RD loss function. Related code: CompressAI/compressai/models/google.py at a4ae2eeef7bdb1b84ba076ac0d650b523f3fa882 · InterDigitalInc/CompressAI · GitHub
3. Ballé, J., et al. (2018). "Variational image compression with a scale hyperprior." arXiv preprint arXiv:1802.01436. Adds a hyperprior (hyper) on top of the basic architecture, related code: CompressAI/compressai/models/google.py at a4ae2eeef7bdb1b84ba076ac0d650b523f3fa882 · InterDigitalInc/CompressAI · GitHub
4. Minnen, D., et al. (2018). "Joint autoregressive and hierarchical priors for learned image compression." Advances in neural information processing systems. Autoregressive (autoregressive) and hyperprior, related code: CompressAI/compressai/models/google.py at a4ae2eeef7bdb1b84ba076ac0d650b523f3fa882 · InterDigitalInc/CompressAI · GitHub

Note: CompressAI can be installed directly via pip on Linux, but does not provide a Windows installation package. Follow the steps below to install:

1. Install the C++ related packages in Visual Studio and restart.
2. Use conda to create a new Python environment and install pytorch, scipy, matplotlib, pytorch-msssim.
3. Activate the above environment, clone CompressAI locally, enter the directory in the terminal, and run pip install .