Low-complexity color demosaicing algorithm based on integrated gradients

Abstract

Color demosaicing is critical for digital cameras, because it converts a Bayer sensor mosaic output to a full color image, which determines the output image quality of the camera. In this work, an efficient decision-based demosaicing method is presented. This method exploits a new edge-sensing measure called integrated gradient (IG) to effectively extract gradient information in both color intensity and color difference domains simultaneously. This measure is reliable and supports full resolution, which allows one to interpolate the missing samples along an appropriate direction and hence directly improves the demosaicing performance. By sharing it in different demosaicing stages to guide the interpolation of various color planes, it guarantees the consistency of the interpolation direction in different color channels and saves the effort required to repeatedly extract gradient information from intermediate interpolation results at different stages. An IG-based green plane enhancement is also proposed to further improve the method's efficiency. Simulation results confirm that the proposed demosaicing method outperforms up-to-date demosaicing methods in terms of output quality at a complexity of around 80 arithmetic operations per pixel.

Bayer Image Demosaicking Using Eight-Directional Weights Based on the Gradient of Color Difference

“…From Table 2, the average values of SSIM show that the image which is interpolated by the proposed method is more similar to the original image. [6], GBTF [8], RI [10], MLRI [12], and MDWI-GF [17]. The magnified part, which is marked by a black box, is shown in Figure 10a.…”

“…The sampled data of the McMaster (McM) images shown in Figure 9 is used for testing. The original sequence of Bayer is B, G, G, R. To demonstrate the effect of the proposed method, the experimental results are compared with DLMMSE [2], IG [6], GBTF [8], RI [10], MLRI [12], and MDWI-GF [17].…”

“…The magnified image can be clearly seen in Figure 11a. The other seven images, which are magnified in the same part as shown in Figure 11, are obtained by DLMMSE [2], IG [6], GBTF [8], RI [10], MLRI [12], MDWI-GF [17], and the proposed method, respectively. Compared with the other six methods, the proposed one is visually more similar to the original one.…”

“…It can be seen that slightly white granular spots which are the distortion appear in the purple parts from Figure 12b-e. Though the proposed algorithm has distortions at the edge of the purple parts, as shown in Figure 12h, the distortion is more serious, visually, as shown in Figure 12b-g. [2], (c) IG [6], (d) GBTF [8], (e) RI [10], (f) MLRI [12], (g) MDWI-GF [17], and (h) the proposed method. [2], (c) IG [6], (d) GBTF [8], (e) RI [10], (f) MLRI [12], (g) MDWI-GF [17], and (h) the proposed method.…”

“…These algorithms estimate adaptive directions by evaluating the similarity among the neighbor pixels. Chung and Chan [6] proposed integrated gradients (IG) to simultaneously extract gradient information from intensity and color differences. Pekkucuksen and Altunbasak [7] suggested an orientation-free edge strength filter, which is used to apply the color difference adaptively in order to provide more edge information.…”