|
福州大学学报(自然科学版) 2017
一种改进的Criminisi图像修复方法
|
Abstract:
针对Criminisi等提出的基于纹理的图像修复方法中存在的修复顺序不准确等问题,提出一种改进的Criminisi图像修复方法. 对优先权函数的计算方法进行改进,引入影响因子,提出新的梯度算法,增强结构性较强地方的修复效果,提高优先权的准确性. 考虑可能存在的误差因素,定义了新的置信度项的更新方式. 实验表明,该算法具有更好的修复效果.
Image restoration is a practical research focus in the field of computer vision. Aiming at the problem of existing repair sequence image restoration method based on exemplar which is proposed by Criminisi and other people,is not accurate,proposes an improved method of repair. The method of calculating the priority function was improved,the introduction of impact factor,propose new gradient algorithm to enhance the effect of the structural repair of strong local and improve the accuracy of priority,consider the possible error factors,define a new term confidence update mode. Experimental results show that the improved algorithm has better repair effect