摘要
目前大部分基于边缘的无参考图像清晰度评价方法中,普遍对严重模糊图像的清晰度不能较好评价.针对这一问题,结合空间域及转换域,提出了基于边缘与小波变换的新的图像清晰度评价方法.该方法在空间域上根据人类视觉系统的特性,由边缘的宽度估计边缘清晰度;并引入多分辨率分解,改善严重模糊图像因检测不到边缘而影响清晰度评价的情况;在转换域上将图像小波分解后每层的所有高频子带合并为一个细节图,并利用该细节图的梯度计算每层的能量,再合并各层能量得到图像的高频能量.最终的图像清晰度由基于空间域的边缘清晰度及基于转换域的高频能量结合得到.实验证明,本文算法得到的客观评价与人类主观评价较一致,性能稳定.
Most of the existing edge-based no-reference image sharpness assessment methods generally do not perform well for heavily blurred images. To solve this problem, a new edged-based and wavelet-based algorithm for estimating image sharpness is proposed, using a combination of spatial domain and transform domain. Considering properties of the human visual system, the method estimates the edge sharpness by measuring the edge width;then the multi-resolution decomposition is adopted to improve the performance when detecting heavily blurred images, which usually influenced by the invisibility of edge. Three high-frequency bands in each layer of wavelet decomposition combine to form a detail map. Each layer＇s energy is computed by the gradient of its detail map and used to get the image high-frequency energy. The final sharpness metric is computed as the product of the edge sharpness and the energy in high- frequency bands. Experiments show that the algorithm is robust and has high correlation with human perception.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第5期1117-1121,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61202294)资助
关键词
图像质量评价
边缘
小波变换
清晰度
无参考
image quality assessment
edge
wavelet transform
sharpness
no-reference
