Weighted Wavelet-Based Spectral-Spatial Transforms For CFA-Sampled Raw Camera Image Compression Considering Image Features
Abstract
To efficiently compress raw camera images captured using a color filter array (CFA-sampled raw images), wavelet-based spectral-spatial transforms (WSSTs) that change a CFA-sampled raw image from an RGB color space into a decorrelated color space have been presented. This study introduces weighted WSSTs (WWSSTs) that work especially for the CFA-sampled raw images with many edges well. The WWSSTs are obtained by considering that each predict step of the conventional WSSTs is constructed by a combination of two 1-D diagonal transforms and by weighting them along the edge directions in the images. The experiments at JPEG 2000-based lossless and lossy compression basically show that compared with the WSSTs, our WSSTs improve the results for the images with many edges by about 0.04 bpp in LBRs, 2.10 [%] in BD-rates, and 0.12 dB in BD-PSNRs while keeping the compression efficiency for the general images.
BibTeX
@inproceedings{icassp2022_weightedwaveletb,
title = {Weighted Wavelet-Based Spectral-Spatial Transforms For CFA-Sampled Raw Camera Image Compression Considering Image Features},
author = {Liping Huang and Taizo Suzuki},
booktitle = {ICASSP 2022},
year = {2022}
}