Abstract: In order to recognize mud spot disease in ancient murals, an algorithm is designed. First, it analyzes textural features of mud spot disease and draws a conclusion based on self-correlation function. Then it analyzes the brightness and colourity based on YcbCr model. Second, it analyzes the whole murals image based on textural features and synthesizes textural features, brightness, colourity each part and obtains textural mask, brightness mask and colourity mask. ?Then it operates them through and-or computing. It gets the mud spot mask. Finally it realizes the automatic segmentation and extraction of the mud spot disease through mud spot mask is?added to the original. It is an effective algorithm in inpainting.

Key words: mud spot disease, textural features, mask, automatic recognition

摘要： 针对唐墓室壁画泥斑病害问题，提出了泥斑病害自动标定算法。用空间自相关函数分析泥斑病害纹理特征，在YCbCr模型下，分析泥斑病害亮度，色度特征；对图像分块处理，分析其每个图像块的纹理、亮度和色度特征，通过阈值分割得到泥斑的纹理、亮度和色度掩码。为了精确标定又提出将泥斑的纹理、亮度和色度掩码进行与或运算，得到泥斑区域掩码，将泥斑区域掩码与原图进行加运算，实现壁画泥斑病害精确标定。通过壁画虚拟修复实验表明这种标定算法不仅标定准确而且提高了壁画虚拟修复效率。

关键词: 泥斑, 纹理特征, 掩码, 自动标定