Research On BAYER Color Filter Array Interpolation Algorithm Based On Sparse Representation

The color image CFA interpolation algorithm interpolates some data to produce an RGB image. Single-chip digital cameras use color filter arrays (CFA) to sample color information, but there is only one color component at each pixel location. In order to obtain the full color image, the missing color components should be interpolated. This process is also called demosaicking. Based on the recent interpolation algorithm, this paper performs the researches on the interpolation algorithm of noise-free CFA image and noise CFA image, mainly including the following three aspects.First of all, this paper introduces the edge adaptive color demosaicking based on the spatial correlation of the Bayer color difference. The correlation of the color channel is extended to the color different channel. In order to prevent interpolation across edges, the edge direction is estimated in different region. And the edge directed interpolation is performed. Experimental results show that the proposed method can prevent interpolation across edges and artifacts.Secondly, this paper introduces the contourlet coefficient to express the image sparsely. At the same time, this paper introduces the local Gaussian model instead of the Laplace distribution as the prior knowledge, in order to utilize the statistics of the neighborhood area. The sparsity of image gradient is added to interpolation process. Experiment results show that the proposed algorithm outperform the classical algorithm in terms of both color peak signal to noise ratio and visual quality.At last, in order to avoid many noise-caused color artifacts in the interpolation process, which are hard to remove in the denoising process. This paper presents a PCA based denoising algorithm, which works directly on the CFA data. The optimal dimension reduction property of PCA is used to reduce noise, and the denoising CFA data is interpolated with the local DCT. Combine two algorithms, this paper proposes the PCA-based spatially adaptive denoising and local DCT interpolation algorithm. Experimental results show that the proposed method is able to improve the vision effect of the reconstruction image.