Image Inpainting Algorithm Based On Adaptive Multiple Dictionaries

Image inpainting is a process that using the known information of the image fills theunkown information, making the recovery image almost close to the visual effect of theoriginal image. Image inpainting has broad applications,including artwork, photographs orimage data, goals and obstacles and the literal of the image. In recent years, sparserepresentation theory has become a hot topic in the field of image processing, therefore thedigital image inpainting algorithm based on sparse representation has the very importantvalue and the significance study. Based on the above analysis, the algorithm of theadaptive multi-dictionaries of the color and grayscale images was studied. Mainlyresearched the following areas.Firstly, using that the adaptive multi-dictionaries can be more effective reconstructingthe color image, an algorithm based on FastICA adaptive multi-dictionaries is proposed torecover the color image. The corresponding sub-dictionaries are trained using theorientation angles and the extracting blocks of the image are classificated. Then, beforethe image block is recovered to use the sparse representation, the best sub-dictionary needto be selected. The training multi-dictionaries overcomes the shortcomings that theindividual dictionary cannot reflect the internal structure features of the image, and thealgorithm enhances the inpainting performance and the adaptivity.Secondly, in order to make full use of the local sparse of image and global sparse，anovel high quality color image adaptive inpainting algorithm jointed local image patchmultiple dictionaries sparse code with the sparse representation analytic contourlettransform or dual tree complex wavelet transform is proposed in this paper. Global sparseis introduced local sparse representation constraint entry, in the algorithm. Theexperimental results show that this algorithm can get a better subjective visual effect andthe repaired image is clearer and has a higher peak value signal-to-noise ratio.At last, for the problem of the blind image inpainting, an image adaptive inpaingtingalgorithm jointed K-SVD multi-dictionaries with PCA is proposed in this pepper. Thealgorithm uses the K-SVD algorithm iteratively updating multi-dictionaries, PCA method to inpaint image, and the median filter is introduced into the mask estimation. Then, Theexperimental results show that the better image inpainting results can be obtained.