针对匹配块不存在的Criminisi算法改进研究

该文针对Criminisi算法在已知区域中不存在合适匹配块而造成的修复效果失真问题,探讨了一种利用匹配块相似度大小,对修复失真区域进行定位和修复的改进算法.算法首先对Criminisi算法在修复过程中所产生的修复失真区域进行定位,然后将该区域分解为结构和纹理两部分,最后分别利用TV模型和纹理合成技术完成对结构、纹理的修复.实验证明,改进算法较好地克服了Criminisi原算法所存在的缺点,有较好的视觉效果.

The Research of Criminisi Improved Algorithm According to the Inexistence of Match Piece

When in the image inpainting process,the Criminisi algorithm is easy to cause inpainting effect distortion because of the inexistence of appropriate match pieces in the known areas.This paper explores a new Criminisi improved algorithm,which makes use of the similarity of match pieces to find out the bad inpainting areas and repairs them again.Firstly,improved algorithm finds out the location of bad inpainting areas.Secondly,it discomposes these areas into structure part and texture part.At last,the algorithm uses TV model and texture synthesis technology to inpaint the structure and texture separately.Experiments confirm that this algorithm overcomes the shortcomings of Criminisi algorithm and has better visual effect.