Remove meat image´s reflective areas with a level set constraint inpainting algorithm based on exemplar patch

Because of moisture existing in meat, there would be reflective areas in the meat image under visible light, which lead to a loss of some muscle information. In order to improve the accuracy for obtaining meat image´s information, it is necessary to restore the reflective areas as much as possible. According to the feature of meat image, combine with the principle of image inpainting based on fast marching method, an improved image completion algorithm by exemplar patch is presented in this paper. First, binding the restoration order with level set distance and replacing Robert operator with Prewitt. Then, by using three factors such as direction, geometric distance and level set distance, known pixels in exemplar patch are weighted during the procedure of matching. In addition, the amount of scanning data is reduced in the process of searching the best matched patch through neighborhood matching. After that, to eliminate staircase effect, add α-cut Mean Filter in the process of inpainting. The experimental results show that the inpainting effect of improved algorithm is better than other algorithms present in this paper, and the method can not only repair the reflective areas of meat image, but also be useful for other images´ inpainting.