Ccd-based Static Image Moir¨¦ Removal Algorithm Research

When there is a regular pattern in the scene, there will be often some inexplicable wave-like stripes existing in the images, which is called colored moire. We often see moire phenomenon in the high frequency parts of the image, it is easier to see moire phenomenon in R/B channels than G channel. Usually, there is often a large area that has colored moire in the image, and it has various formats, so it is difficult to remove colored moire. Now there are three ways to remove colored moire, the first way is to add lowpass filter in front of the camera lens, the second way is to remove colored moire by the posterior image processing software, the last way is to find better interpolating algorithms. In the paper, we focus on the third way, based on the research of the color moire and the discussion of the traditional interpolation algorithms, we bring forward better algorithms.The research in this paper is based on the Bayer CFA with single CCD sensor. According to the structure of the Bayer CFA, it is necessary to interpolate the images to access full color images. We should do the interpolation based on the relativity of the colors and along the texture direction, trying to reduce the probability of spanning the edges.Based on the research of the classical algorithms, we bring forward two improved algorithms. The first algorithm is based on the variation direction of the green pixels, the direction is estimated on the relation value of R and G,B and G.,selecting R or B channel to help estimating the edge direction of G channel.After the preparatory interpolation, we replace the estimated high-frequency by the known values according to the high relation of the three channels, the step restrains the moire phenomenon effectively, and increase the quality of the image at the edge. The second algorithm is based on Hamilton algorithm, adopting different ways to select the direction of the G values, in the interpolation of R/B channel, we add weight value according to the relativity, increasing the validity of R and B channel. We also do improvement of the posterior processing, using bilateral filter to restore the image. At last, we discuss the classical quality assement index, using them to analysis the algorithms in the paper. The experiment proves that the modified algorithms is logical, compared with classical interpolation algorithms, the quality of the images are improved, moire phenomenon has a high degree of removal.