DSSLIC: Deep Semantic Segmentation-based Layered Image Compression
This work addresses image compression for applications like storage and transmission, with potential benefits for tasks such as image search, but it is incremental as it builds on existing deep learning methods.
The paper tackles image compression by proposing a deep semantic segmentation-based layered framework (DSSLIC) that encodes segmentation maps and residuals, achieving better performance than H.265/HEVC-based BPG in PSNR and MS-SSIM metrics across various bit rates.
Deep learning has revolutionized many computer vision fields in the last few years, including learning-based image compression. In this paper, we propose a deep semantic segmentation-based layered image compression (DSSLIC) framework in which the semantic segmentation map of the input image is obtained and encoded as the base layer of the bit-stream. A compact representation of the input image is also generated and encoded as the first enhancement layer. The segmentation map and the compact version of the image are then employed to obtain a coarse reconstruction of the image. The residual between the input and the coarse reconstruction is additionally encoded as another enhancement layer. Experimental results show that the proposed framework outperforms the H.265/HEVC-based BPG and other codecs in both PSNR and MS-SSIM metrics across a wide range of bit rates in RGB domain. Besides, since semantic segmentation map is included in the bit-stream, the proposed scheme can facilitate many other tasks such as image search and object-based adaptive image compression.