On the use of color appearance modeling for efficient compressed-domain image enhancement

In this paper we propose a compressed-domain image enhancement algorithm based on color appearance modeling. We aim at providing an efficient high-quality JPEG color image enhancement algorithm, in terms of color boosting and low computational cost. For that purpose, the high-performance recently proposed iCAM06 model has been adopted, where a wide range of color appearance phenomena are predicted in an efficient manner. In order to reduce the computational complexity, a pseudo-pixel extraction algorithm is applied to the four low-frequency DCT coefficients of each block, resulting in a scaled RGB version of the input image (directly in the DCT domain), where the iCAM06 model is then applied. To cope with the problem of inadequate information of small-sized images, as only 6.25% of the DCT coefficients are used, a 2-fold upsampling is performed at a preprocessing stage. The proposed algorithm outperforms relevant methods, as it is validated by the comparative results.