<ul><li>Continuous improvements in image compression with variational autoencoders have led to competitive learned codecs.</li><li>Quantization during the training process poses challenges due to zero derivatives, requiring differentiable approximations for optimization.</li><li>Proposed method involves retraining parts of the network on quantized latents post end-to-end training for improved accuracy in modeling quantization noise.</li><li>Results show additional coding gain for both uniform scalar and entropy-constraint quantization without increasing complexity, with average savings up to 2.2% in bitrate.</li></ul>

Optimizing Learned Image Compression on Scalar and Entropy-Constraint Quantization

Discover more