Adaptive demosaicing algorithm using characteristics of the color filter array pattern

Abstract

Generally, the color filter array (CFA) image is interpolated considering the correlation between color channels. Previous works first interpolate the green (G) signal, and then obtain the differences between theR/B signal and the reference signal (the initial interpolated G signal). To determine the direction of interpolation, the proposed method computes the horizontal/vertical absolute inter-channel differences directly computed from the CFA image. Then, three color components (R/G/B) are interpolated along the estimated horizontal/vertical directions considering the differences of absolute inter-channel differences. Comparative experiments using 24 test images with six conventional demosaicing algorithms show the effectiveness of the proposed demosaicing algorithm in terms of the peak signal to noise ratio, structural similarity, and subjective visual quality.