形态成分分析是一种基于稀疏模型的图像分析算法, 其中心思想是根据信号组成成分的形态差异性选择两个合适的字典分别用来表示纹理部分和边缘卡通部分, 具有良好的图像修复特性。传统上字典的选择需要由使用者根据图像内容人为确定。提出一种基于图像内容的自适应字典选择方法, 根据最小能量在字典集合中选择最适合当前图像的字典并对图像进行修复。实验证明, 该方法具有良好的图像修复性能。

Morphological component analysis(MCA)is an image analysis method based on sparse model. It can separate overlapping texture and cartoon image layers. Its central idea is to use two adapted dictionaries, one adapted to represent textures, and the other to represent cartoons. MCA is inherent merit performance in image inpainting. In traditional, the dictionaries are selected manually. A method for adaptive dictionary selection against MCA is proposed. The dictionaries are selected according to energy minimization. The images are inpainted according to the dictionaries. The experiments show that the method performs well in image inpainting.