基于清晰度信息与灰度差异熵的图像修复算法

Image Inpainting Algorithm Based on Clarity Information Coupled Gray Difference Entropy

当前较多图像修复算法主要利用图像的结构信息来计算待修复块的优先权,忽略了图像的清晰度信息,导致修复图像中含有振铃以及块现象等问题。对此引入图像的空间频率,设计了基于清晰度信息与灰度差异熵的图像修复算法。首先,将数据项、空间频率和置信度项进行组合,形成待修复块的优先权度量函数,使其在优先权的计算过程中充分考虑了图像的清晰度信息,以提高优先权值计算的准确性,确保合理的修复顺序。然后,采用图像的灰度差异信息,构造灰度差异熵函数,计算合适的样本块大小。最后,联合绝对差平方和SSD与灰度差异熵函数,从图像的色彩差异信息以及灰度差异信息出发,确定待修复块对应的最优匹配块,以修复图像。实验结果表明,相对于已有的图像修复方法而言,所提算法具备更为理想的修复效果,对应的结构相似度值更大,其输出结果中不含有振铃和块现象。

At present, many image restoration algorithms mainly use the structure information of the image to calculate the priority of the blocks to be repaired, while ignoring the clarity information of the image, which results in the defects of ringing and block phenomenon in the repaired image. Based on the spatial frequency model, this paper designs an image restoration algorithm based on sharpness information. Firstly, the data item, spatial frequency model and confidence item were combined to form the priority measure function of the block to be repaired, so that the image definition information was fully considered in the priority calculation process, so as to improve the accuracy and rationality of priority calculation. Then, the gray difference entropy model was constructed by using the gray difference information of the image, and the appropriate size of the sample block was calculated. Finally, the absolute difference square sum model was combined with the gray difference entropy model. Based on the color difference information and gray difference information of the image, the optimal matching blocks corresponding to the blocks to be repaired were obtained accurately to repair the image. The experimental results show that the image restored by this algorithm does not contain ringing and block phenomena, and it has a better visual effect than the image restored by the existing algorithm.