We introduce a stop-code tolerant (SCT) approach to training recurrent convolutional neural networks for lossy image compression. Our methods introduce a multi-pass training method to combine the training goals of high-quality reconstructions in areas around stop-code masking as well as in highly-detailed areas. These methods lead to lower true bitrates for a given recursion count, both pre- and post-entropy coding, even using unstructured LZ77 code compression. The pre-LZ77 gains are achieved by trimming stop codes. The post-LZ77 gains are due to the highly unequal distributions of 0/1 codes from the SCT architectures. With these code compressions, the SCT architecture maintains or exceeds the image quality at all compression rates compared to JPEG and to RNN auto-encoders across the Kodak dataset. In addition, the SCT coding results in lower variance in image quality across the extent of the image, a characteristic that has been shown to be important in human ratings of image quality
@article{arxiv.1705.06687,
title = {Target-Quality Image Compression with Recurrent, Convolutional Neural Networks},
author = {Michele Covell and Nick Johnston and David Minnen and Sung Jin Hwang and Joel Shor and Saurabh Singh and Damien Vincent and George Toderici},
journal= {arXiv preprint arXiv:1705.06687},
year = {2017}
}