实时图像质量严重影响景象匹配性能,但目前尚无通用可行的评价标准来衡量其质量,而且不同匹配算法对各种图像特征的敏感性有差异,需要根据算法对图像质量评价问题展开分类研究。文中首先给出了图像质量评价的工作方式,然后从成像质量和目标凸显程度两个方面进行指标分析,提出了包括信噪比、模糊度、目标局部凸显程度和全局凸显程度在内的指标体系。实验表明,各指标与匹配性能之间具有单调一致关系。

The real time image quality influences the performances of scene matching extensively, but there is no general metric for its quality assessment. Furthermore, different matching algorithms are of diverse sensitivity to various image features. Therefore, the study on image quality assessment (IQA) should be carried out according to the style of algorithms. Firstly, applications of IQA were described. Then, the assessment criteria were analyzed from two aspects: imaging quality and target saliency metric. An evaluation system, which includes signal to noise ratio (SNR), blur metric, local saliency metric (LSM) and global saliency metric (GSM), was presented. Experiments show that these criteria are consistent with the probability of the matching algorithm.