改进的GrabCut算法在古代壁画分割中的应用

Application of Improved GrabCut Algorithm in Ancient Mural Segmentation

壁画作为一种珍贵的文化遗产,在经历数千年的沉淀后普遍存在毁坏破损的现象,其保护工作刻不容缓.利用智能信息处理技术对壁画进行自动分割是壁画数字化保护的一个重要组成部分.针对壁画噪声明显、边缘不清晰的特点,提出了一种融合小波去噪和边缘增强的改进的GrabCut算法新模型.该算法用小波变换对壁画图像进行分解,并采用自适应特征阈值方法去除壁画图像中的噪声,然后融合Sobel算子和Canny算子提取壁画轮廓以增强边缘,在此基础上对壁画图像进行分割.仿真实验通过分割效果和PSNR,Kappa,Error这3个指标来评价本文算法模型,实验结果表明:本文算法相对于对比方法的PSNR值平均提高了7.817,Kappa值平均提高了0.076,Error值降低了0.080,说明本文提出的模型在分割含有噪声、边缘模糊的壁画图像时不仅具有良好的抗噪能力,而且分割效果更好,准确度更高.

As a kind of traditional cultural heritage,the existing murals are generally damaged after thousands of years of precipitation.Automatic mural segmentation using intelligent information processing technology is an important part of digital protection for murals.In view of the obvious noise and blurred edges of existing murals,a new improved GrabCut algorithm model based on wavelet denoising and edge enhancement was proposed for murals.The proposed algorithm used wavelet transform to decompose the mural image,and adopt the method of adaptive feature threshold to remove the noise in the mural,and then combined Sobel operator and Canny operator to extract the contour to enhance the edge.On this basis,the mural image was segmented.The simulation experiment evaluated the algorithm model through segmentation effect and three indexes:PSNR,Kappa and Error.Experimental results show that compared with the related method,the average PSNR value of the proposed algorithm increases by 7.817,the average Kappa value increased by 0.076,and the Error value is reduced by 0.080,which fully illustrates that the proposed method not only has good anti-noise ability,but also has better segmentation effect and higher accuracy especially when it is used to segment mural images with noise and blurred edges.