摘要
在景象匹配中,为了对匹配算法的可靠性和图像的适用性进行分析,必须对图像质量进行评价。景象匹配中的图像质量评价主要分析图像目标的可匹配性,可归结为目标的凸显性。通过将边缘结构相似度算法运用于小波域多尺度图像计算不同尺度图像的疑似目标数目,并对各级图像的疑似目标数目进行加权,获得最终疑似目标数目,设计了目标凸显性指标。经实验分析,该指标与图像匹配概率一致,符合景象匹配的评价需求。
Image quality assessment for scene matching is necessary to analyze the robust of matching algorithm and applicability of image. Image quality assessment for scene matching aims at targets matching probability which comes down to target saliency. This paper computed the number of suspicious targets in wavelet domain by computing edge structure similarity between target template and different domain of real-time image. The final number of suspicious targets was the sum of the number of suspicious targets in different level of wavelet transform with different weight. Then, designed an evaluation index based on the final number of suspicious targets to represent target saliency. Experiments show that the evaluation index is consistent with matching probability and effective for scene matching.
出处
《计算机应用研究》
CSCD
北大核心
2012年第3期1152-1154,共3页
Application Research of Computers
关键词
目标凸显性
图像质量评价
边缘结构相似度
景象匹配
target saliency
image quality assessment
edge structure similarity
scene matching
