基于Criminisi算法的图像修复顺序的研究

Research on Image Inpainting Order Based on Criminisi Algorithm

Criminisi算法是用于修复大尺度破损图像的经典算法。由于该算法中像素块的置信度会随着修复的进行迅速降为零,并且也存在数据项为零的可能,而置信度项和数据项中任一项为零都会导致错误的修复顺序,造成错误信息的累积,最终影响图像的修复效果。因此,文章改进了优先权的计算公式,并用实验验证了改进后优先权计算公式的优越性,从而在一定程度上提高了图像修复后的视觉效果。

The Criminisi algorithm is used to repair the large scale damage image, it is a classic algorithm.However, the confidence term of the algorithm will be reduced to zero with being repaired, and the data item may be zero. The confidence level and data items in any one of zero will result in an error of repair order, causing cumulative error messages, and ultimately affecting the repair effect of the image. Therefore, this paper improves the formula of priority, and experimentally verify the superiority of the new priorities of the formula, to a certain extent, improve the image restoration of visual effects.