The Research On Fast Image Inpainting Algorithms

Digital image inpainting has been a very important basic research subject, it belongs to an ill-posed problem from mathematical theory. At present, though some image inpainting algorithms have got good repair effect, their time consuming is very serious. This paper mainly studies the structure image fast inpainting and texture image fast synthesis restoration techniques, the main research work is as follows.-For small scale damaged structure image repair, and in order to make full use of the local area information of image to speed up information diffusion effectively and overcome the inefficiency of existing algorithms in the protection of image texture details when repairing, a novel Iterative filtering fast image inpainting algorithm is proposed in this paper. The proposed algorithm first gets a group of points to be repaired in an order from outside to inside of damaged area, then, the Gaussian mask exposure areas (field areas) of points to be repaired are selected adaptively according to their positions; Next, the smooth strength of filtering process is accurately controlled by adjusting the Gaussian kernel weight coefficients with the directions of Isophotes of field points and the distances of them relative to the point to be repaired.For large scale damaged texture image repair, based-on researching texture synthesis image inpainting algorithms, this paper puts forward an adaptive texture synthesis image inapainting algorithm(ATSII). This algorithm adjusts adaptively the size of target block according to its texture component information, reducing the iterative repair times of target blocks for the structure’image repairing, and it improves the priority function, to some extent, improving its robustness. The algorithm reduces the searching time in a great degree when searching the match blocks of the target ones by restricting the search area according to the principle of image connectivity. In addition, for the calculation method of best match blocks, their distance information and texture characteristic similarity are considered in to enhance the accuracy for the optimal matching selections.Based on wavelet transform multiresolution decomposition, high frequency coefficients’characteristics, and strong coefficient correlation etc., in the end, a wavelet domain texture synthesis fast image restoration algorithm is proposed by utilizing the algorithm steps of ATSII in this paper. This algorithm firstly utilizes high frequency coefficients instead of gradient to repair low frequency coefficients using ATSII, after a target block is repaired, according to the principle of wavelet strong coefficient correlation, the damaged areas in other tree high frequency components will be repaired by the same repair vector. Because of the damaged area is narrowed by wavelet transform, the algorithm further shortens repair time in a great degree compared with ATSII.The experimental result shows that the proposed algorithms in this page can get better subjective visual effects and higher repair efficiency than the similar algorithms, and the repaired images are clearer. In short, they obtain better repair effects.