摘要
Criminisi图像修复算法用于优先权计算时,对图像纹理和结构信息的处理存在不足,极易导致错误的匹配和纹理延伸。针对此问题,提出基于色阶阈值模型的Criminisi图像修复算法。首先,在优先权计算中结合修复块的色阶信息建立色阶阈值模型,据此模型获取优先修复块,用以增加优先权的可靠性,避免纹理延伸和结构混乱。然后,在搜寻最佳匹配样本块时,以加入样本块的色度和亮度信息作为比较对象,以便在已知像素块中更加快速准确地搜寻出最佳匹配样本块,减少匹配错误。与其他图像修复算法相比,该算法的修复顺序更加稳定,修复效果满足视觉要求。
In view of the lack of reasonable consideration of texture and structure information in the priority calculation of criminisi image inpainting algorithm,which is prone to wrong matching and texture extension,an improved criminisi image inpainting algorithm based on the color scale threshold model is proposed in this paper.In order to increase the reliability of the priority and avoid the phenomenon of texture extension and structure confusion,a color level threshold model is established based on the color level information of the repair block.Then,when searching for the best matched sample block,the information of chroma and brightness added in the sample block is taken as the comparison object in order to search out the best matched sample block faster and more accurately in the known pixel block,and reduce the occurrence of matching errors.Compared with different image inpainting algorithms,the results show that the proposed method is more stable in the order of restoration,and the effect of restoration is more in line with the visual requirements.
作者
赵卫东
秦锋
ZHAO Weidong;QIN Feng(School of Information Engineering,Chuzhou Vocational and Technical College,Chuzhou Anhui 239000,China;School of Computer Science,Anhui University of Technology,Ma'anshan Anhui 230009,China)
出处
《重庆科技学院学报（自然科学版）》
CAS
2020年第4期70-75,共6页
Journal of Chongqing University of Science and Technology：Natural Sciences Edition
基金
安徽省职业与成人教育学会2018年教育科研规划项目“职业院校优质教学资源库共建共享机制研究”(AGZ18071)
滁州职业技术学院重点应用研究项目“大数据环境下基于数据挖掘技术的教科研管理系统的优化与改进”(YJZ-2019-43)
安徽省大规模在线开放课程(MOOC)示范项目“web标准网页设计”(2017MOOC294)。
