基于图像分解与字典分类的单幅图像去雨算法
发布时间：2018-08-19 20:43
【摘要】：针对单幅图像下,基于稀疏表示的去雨算法存在残差较大而导致图像恢复效果不理想的问题,提出了一种优化图像高频部分几何分量的去雨方法.首先采用平滑滤波做图像分解,得到雨图像的高频部分;然后结合稀疏表示与近邻传播算法分离出图像高频部分的雨分量,用图像的高频部分减去雨分量并做平滑处理,以此作为几何分量;此外,对稀疏表示过程得到的字典进行再分类,完善雨分量与非雨分量的区分,最后完成图像恢复.实验结果表明,该方法能有效利用图像的几何信息来解决纹理恢复误差较大的问题,实现更精确的纹理恢复和雨分量去除.
[Abstract]:Aiming at the problem that the rain-removing algorithm based on sparse representation has large residuals which lead to unsatisfactory image restoration effect, a rain-removing method is proposed to optimize the geometric components of the high-frequency part of the image. Firstly, the high frequency part of rain image is obtained by using smoothing filter, then the rain component of high frequency part of image is separated by sparse representation and nearest neighbor propagation algorithm, and the rain component is subtracted from the high frequency part of image and smoothed. In addition, the dictionary obtained from the sparse representation process is reclassified to perfect the distinction between rain component and non-rain component, and finally the image restoration is completed. The experimental results show that the method can effectively use the geometric information of the image to solve the problem of large texture recovery error and achieve more accurate texture recovery and rain component removal.
【作者单位】： 天津大学电气自动化与信息工程学院;
【基金】：国家自然科学基金资助项目(61472274);国家自然科学基金重点项目(61632081)~~
【分类号】：TP391.41
[Abstract]:Aiming at the problem that the rain-removing algorithm based on sparse representation has large residuals which lead to unsatisfactory image restoration effect, a rain-removing method is proposed to optimize the geometric components of the high-frequency part of the image. Firstly, the high frequency part of rain image is obtained by using smoothing filter, then the rain component of high frequency part of image is separated by sparse representation and nearest neighbor propagation algorithm, and the rain component is subtracted from the high frequency part of image and smoothed. In addition, the dictionary obtained from the sparse representation process is reclassified to perfect the distinction between rain component and non-rain component, and finally the image restoration is completed. The experimental results show that the method can effectively use the geometric information of the image to solve the problem of large texture recovery error and achieve more accurate texture recovery and rain component removal.
【作者单位】： 天津大学电气自动化与信息工程学院;
【基金】：国家自然科学基金资助项目(61472274);国家自然科学基金重点项目(61632081)~~
【分类号】：TP391.41
【相似文献】
相关期刊论文 前10条
1 孟勇,洪丹辉,毛丹;测度熵在图像纹理分析中的应用[J];计算机应用与软件;2000年08期
2 吴涛;秦昆;;图像纹理特征数据挖掘的理论与方法探讨[J];计算机时代;2006年08期
3 方玲玲;王相海;;图像挖掘研究[J];计算机科学;2009年08期
4 高振宇;杨晓梅;龚剑明;金海;;图像复杂度描述方法研究[J];中国图象图形学报;2010年01期
5 刘勇,施万昌,徐玉兰;图像差异的分析与识别[J];复旦学报(自然科学版);2000年05期
6 罗l，
本文编号：2192779
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2192779.html
