Accelerated Deep Lossless Image Coding with Unified Paralleleized GPU Coding Architecture
This work addresses the computational bottleneck for real-world applications of learned lossless image compression, with incremental improvements in speed and domain-specific adaptability.
The authors tackled the problem of slow learned lossless image compression by introducing a parallelized GPU architecture, achieving encoding/decoding times under one second for grayscale images and beating FLIF in compression rate.
We propose Deep Lossless Image Coding (DLIC), a full resolution learned lossless image compression algorithm. Our algorithm is based on a neural network combined with an entropy encoder. The neural network performs a density estimation on each pixel of the source image. The density estimation is then used to code the target pixel, beating FLIF in terms of compression rate. Similar approaches have been attempted. However, long run times make them unfeasible for real world applications. We introduce a parallelized GPU based implementation, allowing for encoding and decoding of grayscale, 8-bit images in less than one second. Because DLIC uses a neural network to estimate the probabilities used for the entropy coder, DLIC can be trained on domain specific image data. We demonstrate this capability by adapting and training DLIC with Magnet Resonance Imaging (MRI) images.