Adaptive Region Demosaicking Using Multi-Gradients

Abstract Due to the cost and the physical structure of color image sensor, most modern digital cameras acquire images using a single image sensor overlaid with a color filter array. At the location of each pixel only one color sample is taken, so a process called color demosaicking is required to reconstruct the full color images. General interpolation method may blur the texture and introduce color artifacts near edges. An adaptive region demosaicking using Multi-gradients is put forward in this paper to reduce the color artifacts. We first compute green according to the gradient operators and the weighted average method. The chrominance channels are then reconstructed base on the constant-hue rules. Finally, the result is optimized by the correlation of the detail. Stimulation studies indicate that the algorithm can improve the PSNR, sharpen texture and edge of the image and enhance image quality.