微小零件显微检测图像的清晰度评价

Sharpness Evaluation of Microscopic Detection Image for Micro Parts

摘要:

根据微小零件显微检测图像纹理相似、边缘信息太少和灰度分布范围有限等特点,从算法的基本原理出发,对传统清晰度评价函数进行了分析。针对传统清晰度评价函数无法兼具高灵敏度和抗噪性的缺陷,提出了一种基于局域方差信息熵的清晰度评价算法。该算法利用局域方差对各灰度级的自信息量进行加权,一方面弥补了信息熵函数缺失的空间信息,避免清晰度误判；另一方面增加清晰区域像素参与计算信息量时的权重,同时减少背景区域和噪声区域像素计算的权重,从而提高函数灵敏度。实验结果表明,与传统清晰度评价函数相比,局域方差信息熵函数不仅灵敏度很高,而且具有较好的抗噪性,可用于实际的自动对焦系统中。

Abstract:

According to the characteristics of micro parts microscopic detection image, including the image texture is similar, the edge information is too little and the gray distribution range is limited, the traditional sharpness evaluation function is analyzed based on the basic principles of algorithm. Aiming at the defect that the traditional sharpness evaluation function can not have own high sensitivity and noise immunity, an algorithm based on local variance information entropy is proposed. The method uses the local variance to weight the self-information of each gray level,on the one hand, it makes up for the lack of spatial information of information entropy and avoids misjudgement of sharpness; on the other hand, it can increase the weight of clear region pixels when they participate in the calculation of information, while reduces the weight of background and noise region pixels, thereby improves the function sensitivity. The experimental results show that compared with the traditional sharpness evaluation function, the local variance information entropy function not only has high sensitivity, but also has better noise immunity and is suitable for actual auto-focusing systems.