Based On The Bayer Color Filter Array Interpolation Algorithm

In order to reduce cost and complexity, digital camera manufacturers usually employ a single CCD or CMOS sensor whose surface is covered with a color filter array, named as CFA. The CFA limits each sensor pixel just sampling one of the three primary color values (red, green or blue). The most popular CFA is Bayer CFA. In order to get the full color image, the other two missing color samples must be interpolated. This process is commonly referred as demosaicing. The former literatures have described a large number of effective algorithms, ranging from simple linear interpolation to sophisticated adaptive ones. Though adaptive algorithms are believed to outperform non-adaptive ones in terms of the quality of the results, but they are difficult for the hardware implement.In this paper, we first review some commonly used algorithms and evaluate their superiority and deficiency according to the testing results of some typical images. Then propose an improved demosaicing algorithm for the Bayer CFA based on hardware implement, which introduces a direction flag for each pixel, and a smooth method for detecting edge direction. It is easy to implement and can be used in real time on high speed cameras in practice, but with low quality. So, we propose another novel algorithm using the cubic spline interpolation along the relevant direction and computing weights by trigonometric function of the gradients. The algorithm also adopts the correlation between color channels. And it proved highly effective based on some tests.In order to assess the performance of the proposed methods, we propose three types of image measures to quantify the performance compared to some other methods. The first type evaluates the fidelity of the reconstructed image by computing the peak signal-to-noise ratio and Euclidean distance in CIELAB color space between the original and reconstructed image in smooth and edge regions respectively. The second type accounts for a major artifact—zipper effect. The third type assesses the cost of the hardware implement and the feasibility. Last, we give out the results of visual comparison and quantitative comparison based on some test images.