An Efficient Algorithm for Luminance Optimization in Chroma Downsampling

An Effective Bicubic Convolution Interpolation-Based Iterative Luma Optimization for Enhancing Quality in Chroma Subsampling

“…Based on the Kodak dataset with 24 images, the IMAX dataset with 18 images, the Video dataset with 200 images, the CI dataset with 8 images, and the SCI dataset with 20 images, the quality, execution time, and quality-bitrate tradeoff improvements of our iterative luma modification method are demonstrated when compared with the two state-of-theart methods [4], [11]. For convenience, the five considered chroma subsampling methods are denoted by the set C s = {4:2:0(A), 4:2:0(L), 4:2:0(R), 4:2:0(D), Anchor}; the three considered chroma upsampling methods are denoted by the set C u = {NN, 4-tap, INTER_CUBIC} where the nearest neighbor (NN) method and the 4-tap method are supported by the VVC platform [18].…”

“…Although the search interval is short for the pixel in the smooth area, it may be long for the pixel in the textural area. Therefore, applying a differentiation technique on the search interval [4], Lin et al [11] proposed a luma optimization (LO) method to reduce the number of luma values examined in the search interval. However, due to the differentiation technique used, the luma values examined in [11] are real values, so it still needs to take a floor operation and a ceiling operation on each examined luma value, and then a better luma modification solution is selected.…”

“…Relative to the current two methods [4], [11], the three contributions of the proposed nonlinear interpolation-based, namely the bicubic convolution interpolation (BCI)-based, iterative luma modification method for I RGB are clarified as follows.…”

“…Differing from the COPY-based pixel-distortion function [4], [11], we propose a new and more effective BCI-based pixeldistortion function. Further, the convex property of the proposed distortion function, which serves as the base of the initial luma modification solution, is proved.…”

“…Based on the Kodak [10], IMAX [7], SCI (screen content images) [14], CI (classical images) [20], [21], and Video [17] datasets, the thorough experimental results have demonstrated that on the newly released Versatile Video Coding (VVC) platform VTM-12.0 [18], our BCI-based iterative luma modification method achieves substantial quality, execution time, and quality-bitrate tradeoff improvements when compared with the LM method [4] and the LO method [11]. The considered quality metrics include the CPSNR (color peak signal-to-noise ratio), SSIM (structural similarity index measure) [19], and visual effect.…”