摘要
针对基于样本的图像修复技术在修复井下煤岩图像时存在纹理过渡延伸和边缘结构不连续的问题,提出一种强化边缘结构的分段自适应图像修复算法,其在Criminisi算法中引入基于ISEF的数据项和等照度线曲率信息构成新的优先权函数,解决了修复顺序不当的问题;并利用局部方差特征与信息熵划分的区域类型自适应选择样本块大小,解决了边缘保持的问题。在常用测试图像与孟村煤矿的煤岩图像上进行了修复实验,相较于传统的Criminisi方法,该算法的平均PSNR分别提升了0.37 dB与1.33 dB,平均SSIM分别提升了0.0023与0.0027。实验结果表明,该算法对纹理结构信息复杂的图像具有较好的修复效果,为煤壁图像的修复奠定了基础。
Aiming at the problems of texture transition extension and edge structure discontinuity in the repair of underground coal rock images by exemplar-based image inpainting techniques,this paper proposed a piecewise adaptive image inpainting algorithm with enhanced edge structure.The algorithm introduced ISEF-based data items and isoillumination line curvature information into the Criminisi algorithm to build a new priority function,which solved the problem of improper inpaint order.Then,it adaptively selected the sample block size by using the region types divided by the local variance characteristics and information entropy,which solved the edge preservation problem.Finally,it conducted inpainting experiments on commonly used test images and coal rock images of the Mengcun coal mine.Compared with the traditional Criminisi method,the average PSNR of the algorithm were increased by 0.37 dB and 1.33 dB,and the average SSIM were increased by 0.0023 and 0.0027,respectively.The experimental results show that the algorithm has a good inpainting effect on images with complex texture structure information,which lays a foundation for the restoration of coal wall images.
作者
吕伏
张文丽
Lyu Fu;Zhang Wenli(School of Software,Liaoning Technical University,Huludao Liaoning 125105,China;Dept.of Basic Education,Liaoning Technical University,Huludao Liaoning 125105,China)
出处
《计算机应用研究》
CSCD
北大核心
2023年第6期1900-1905,共6页
Application Research of Computers
基金
国家自然基金青年基金项目(51904144)
国家自然基金资助项目(51874166,51974145)。
