Image Inpainting Based FMM Algorithm

For all kinds of reasons, caused by the human or the nature, digital photos and film videos may be destroyed with several regions losing their pixels. These destructions may destroy the integrity of the data. Then digital image in-painting technology came out to solve this problem. By using the remaining information and one algorithm or one process, we can repair these pixels and it gives a good visual efforts.This paper first introduced the history background and the applications of image in-painting. Image in-painting is now widely used in many fields, not only to repair the destroyed region of the image, but also applied on object remove, words remove, image compressing, super-resolution analysis and other applications. Finally we summarize the research of others researchers.Then we introduced the mathematical model and characteristics of digital image in-painting techniques and the mathematical theory. Compared with the result made by the others methods, we at last summarized the advantages and disadvantages of method based the FMM algorithm. The FMM method can spend less time in the computing and contribute a good global effort. But due to the theory itself, it is weaker on edge maintaining.To overcome the disadvantage of FMM method we tried two points to improve the result effort:First, changing the dir index to a pixels selecting process. We concluded the gradient information into a gradient matrix (contained the amplitude and the direction), by using the4neighbor pixels to compute. Secondly, in the traditional FMM method, the in-painting sequence is just from the outside to the inside, without the detail defined. In fact, the confusion of the in-painting also may add the noise to the process. So we used edge detection and the gradient matrix to find the edge breakpoint, and then use the gradient information to predict the edge in the in-painting region. By these edges we can define the further in-painting sequence, and reduce the influence made by a disordered in-painting sequence.From the experiment result, we found that comparing with the traditional FMM method, the improved FMM method could contribute a better edge effort and remain the advantage of less time-consuming at the same time.