Learned Image Compression

Project Title

Learned Image Compression Using Discretized Gaussian Mixture Likelihoods and Attention Modules

Description

Implemented a deep learning based image compression model using discretized Gaussian Mixture Models for precise entropy estimation and attention modules to improve visual quality. Achieves state-of-the-art compression performance with fewer bits and better image reconstruction, matching or outperforming standards like JPEG, JPEG2000, HEVC, and even VVC in PSNR and MS-SSIM.

Citation

Please cite the following papers if you use this work:

@inproceedings{cheng2020image,
    title     = {Learned Image Compression with Discretized Gaussian Mixture Likelihoods and Attention Modules},
    author    = {Cheng, Zhengxue and Sun, Heming and Takeuchi, Masaru and Katto, Jiro},
    booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
    year      = {2020}
}

License

The license for this project can be found here: MIT