文档详情

纹理损失最小约束下的跟踪图像阴影去除算法的改进.doc

发布:2017-02-28约7.61千字共11页下载文档
文本预览下载声明
纹理损失最小约束下的跟踪图像阴影去除算法的改进   摘 要: 图像采集过程中由于光照角度的影响产生阴影,对图像阴影的有效去除能提高图像的暗原色修复能力,改善成像质量。传统的阴影去除方法采用盒子滤波算法,在图像出现光照色差的情况下,阴影去除的效果差。提出一种基于纹理损失最小约束的跟踪的图像阴影去除算法。对图像纹理分块结构进行重构,获取图像阴影的暗原色特征,对图像进行降噪处理,构建纹理损失最小约束函数,以此为约束参量图像阴影跟踪自适应均衡补偿,实现阴影有效去除。仿真结果表明,该算法进行图像阴影去除的效果较好,峰值信噪比较传统方法有所提高,展示了较好的图像处理性能。   关键词: 纹理损失; 最小约束; 图像阴影去除; 峰值信噪比   中图分类号: TN911.73?34; TP391.41 文献标识码: A 文章编号: 1004?373X(2016)24?0104?05   Improvement of image shadow tracking and elimination algorithm based on   texture loss least constraint   YAN Feng, ZHANG Jin, WU Shandan   (Vocational and Technical College of Inner Mongolia Agricultural University, Baotou 014109, China)   Abstract: As the box filtering algorithm adopted in the traditional shadow elimination method has poor shadow elimination effect when the illumination color aberration appears in the image, an image shadow tracking and elimination algorithm based on texture loss least constraint is proposed, with which the texture block structure of the image is reconstructed to obtain the dark primary color characteristics of the image shadow, and then the image is processed with noise reduction. The texture loss least constraint function is constructed, which is taken as a constraint parameter to perform the image shadow tracking adaptive equalization compensation and eliminate the shadow effectively. The simulation results show that the algorithm has good effect of image shadow elimination, the peak signal?to?noise ratio (SNR) is better than that of the traditional method, and has the superior image processing performance.   Keywords: texture loss; least constraint; image shadow elimination; peak signal?to?noise ratio   0 引 言   随着现代信息处理和光学处理技术的发展,数字图像处理技术得到长足的进步,数字图像处理而今广泛应用在人们的日常生活摄影、军事目标识别、地质环境监测等各个领域,并展示了较好的应用前景。图像在多重光融合背景下进行采集过程中,由于受到光照的方向和强度的干扰和影响,导致白平衡失真和偏差,从而产生图像阴影,图像阴影中暗原色的存在影响图像的纹理识别性能,在图像目标识别和精密图像纹理分析中影响应用性能,因此,需要进行图像阴影去除方法的研究,提高图像分析和处理的能力,相关的算法研究受到人们的重视[1]。多重光融合图像在图像成像和采集过程中受到扰动较大,造成图像阴影,阴影部位的图像会导致部分信息丢失,传统方法中,对图像阴影的去除方法主要有基于颜色空间特征分解的
显示全部
相似文档