Colorization-based image coding is a new color image compression technique which compresses the color information of a color image into a small set of representative pixels (RP). The main issue in colorization-based coding is to design good RP extraction and colorization methods such that the compression rate and the quality of the reconstructed color become good. In this paper, we propose an iterative local regression method for RP extraction which takes the major group priority into account. It is shown that the problem of computing the RP can be formulated into a local regression problem. The local regions are obtained by performing a hierarchical image segmentation on the luminance channel both in the encoder and the decoder. Then the local regression parameters, i.e., the parameters that express the local relationship between the pixel values in the luminance channel and the chrominance channels are computed by taking the hierarchical structure into account. At the initial stage, the local regression parameters are calculated to fit the major tendency of the pixels in the segmented regions of the largest scale in the hierarchical structure. After that, additional regression parameters are extracted to reduce the remaining residual errors between the original color and the colors reconstructed by the previously computed local regression parameters. By repeating these steps, an RP set is extracted which can colorize the image in the decoder with sufficient detail. Experimental results show that the proposed algorithm performs better than the JPEG standard and conventional colorization-based coding methods in terms of the compression rate, PSNR and SSIM values.
Bibliographical noteFunding Information:
This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning (Nos. 2015R1A2A1A14000912 and 2013R1A1A4A01007868 ).
© 2016 Elsevier Inc.
All Science Journal Classification (ASJC) codes
- Signal Processing
- Computer Vision and Pattern Recognition
- Statistics, Probability and Uncertainty
- Computational Theory and Mathematics
- Electrical and Electronic Engineering
- Artificial Intelligence
- Applied Mathematics