Image Inpainting Algorithm Based on Saliency Map and Gray Entropy

Image inpainting algorithms based on separated priority are easily misled by image texture information, have poor accuracy in searching for matching patches with high priority and often result in inconsistent texture propagation and edge structure. Additionally, it is difficult to obtain the best-matching patch within a fixed range based on only color information. By considering the attention point of human vision and the statistical information of an image, an image inpainting algorithm based on saliency mapping and gray entropy is proposed. A saliency map is added to the priority stage, which ensures that the parts with strong structural information and visual importance are completed preferentially. The best-matching patch is determined by comprehensively considering the color information and saliency features. The search range of the matching patch is adaptively controlled based on gray entropy. Experiments concerning scratch damage, text removal and large area object removal are compared. The results of the proposed method have better visual effects and are superior in regard to the consistency of the edge structure and texture. The efficiency is similar to methods with a fixed local search range. The objective evaluation results also validate the performance of the proposed method.