Image inpainting using gradient features and color consistency

In existing exemplar-based algorithms, the confidence value tends to reach zero rapidly, filling order is unstable, and mismatch can easily occur easily. To address these problems, an image inpainting algorithm using gradient features and color consistency was proposed. To obtain a more stable filling order, mean gradient was introduced to represent the change characteristics of an image structure and texture. Mean gradient was also applied to calculate the priority to ensure that the structure was preferentially populated and the texture information was appropriately extended. A confidence update term based on an S-shaped function was also proposed to avoid rapid decay of the confidence term. Color consistency between the candidate patch and the target patch was also combined with color information to identify the most similar patch, and to reduce the false matching rate. Experimental results demonstrate that the peak signal-to-noise ratio of the proposed algorithm is at least 0.82 dB higher than that of existing algorithms, which indicates the validity of the proposed method. The results also show that the proposed algorithm can make the inpainting order more stable and can reduce the error matching rate, which brings the repaired images more in line with human visual requirements. © 2019, Science Press. All right reserved.