摘要
提出了一种新的图像增强质量客观评价算法。算法综合考虑图像局部区域的背景平均亮度和空间复杂度对视觉分辨力的影响,导出了用于判断局部灰度级跃变是否被有效感知的重要参数——临界可见偏差JND(Just Noticeable Difference),并利用此参数分别计算出图像增强前后有效感知像素数的变化情况,从而定量地给出图像增强质量的评价结果。测试结果表明,本文算法的性能优于传统的平均局部方差法,且评价结果与主观评价结果吻合。
A novel objective quality evaluation algorithm for enhanced images is presented. The proposed algorithm comprehensively considers the effects of loeal area average brightness of the baekground and spaee complexity on visual resolution. In order to judge whether the varianee of gray is effeetively pereepted or not,the just notieeable differenee（JND） is derived. The ehanges in the number of pereeptible pixels between the original image and the enhaneed image can be caleulated by use of JND parameter. Thus quantitative results to assess the quality of enhaneed images are obtained. The experimental results demonstrate that the proposed algorithm has better performanee than the conventional average loeal varianees method （ALV） ,and assessment results are consistent with those of the subjective perceptual method.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2008年第2期258-262,共5页
Journal of Optoelectronics·Laser
基金
国家自然科学基金资助项目(10704043)
南开大学科技创新基金资助项目
关键词
图像质量评估
图像增强
视觉感知
临界可见偏差
image quality evaluation
image enhancement
human visual perception
Just Notieeable Differenee
