Rate-distortion optimized image compression based on image inpainting

Inspired by recent advancements in image inpainting techniques, an image coding framework is proposed in this paper. In the framework, an original image is analyzed at the encoder side such that a number of the regions are skipped intentionally. A drop map is extracted and compressed into the generated bit stream to indicate the skipped regions. The image is recovered through the inpainting process by taking advantage of the available portion of the decoded image and the drop map at the decoder. Furthermore, the rate-distortion optimization is introduced to select the blocks to be removed for the better performance. Only the blocks containing certain visual features and satisfying rate-distortion criterion are dropped. A practical system is constructed to verify the effectiveness of the compression approach. Evaluations have been made in comparison with baseline JPEG and H.264/AVC. Compared to the baseline JPEG, the proposed algorithm obtains obvious visual quality improvements, as well as PSNR gains. Moreover, the proposed algorithm outperforms H.264/AVC intra coding under the low bit rates.