检焦系统中基于改进Sobel算子的图像清晰度评价函数研究

Research on image clarity evaluation function based on improved Sobel operator in focusing system

针对检焦系统中现有基于图像处理的检焦算子灵敏度不足、抗噪性能弱等问题,该文结合人类视觉系统(HVS),提出了采用灵活性阈值的多方向权重Sobel算子的改进图像清晰度评价方法。该方法将原本仅有两个梯度卷积模板的Sobel算子增加到8个方向;结合HVS特性设置各方向算子卷积的权重值;采用多诺霍收缩阈值为算子预设阈值。实验结果表明,该文提出的改进Sobel算子不仅使图像清晰度评价值更加全面可靠,而且改善了原来阈值灵活性差的问题。与传统清晰度评价方法相比,该算子平均灵敏度因子提升了0.086~0.215,在保证检焦范围的同时提升了曲线灵敏度。在噪声影响下,该方法有效减少了传统方法中出现的局部极值点个数,保持了良好的稳定性,改善了传统清晰度评价方法抗噪性弱的缺陷。

Aiming at the problems of low sensitivity and poor anti-noise performance of existing image processing based focusing algorithms,this paper proposes an improved imag e clarity evaluation method combined with the multi-direction weight Sobel operator and flexible threshold,which is based on the Human Visual System(HVS).This method increases the Sobel operator with only two gradient direction templates to 8 directions;then sets the weight value of the convolution of each direction operator according to the HVS characteristics;adopts the Donoho shrinking threshold as the operator preset threshold.The experimental results show that the improved Sobel algorithm proposed in this paper not only makes the image sharpness evaluation value more comprehensive and reliable,but also improves the poor flexibility of inherent threshold.Compared with the traditional sharpness evaluation method,the average sensitivity factor of this algorithm is increased by 0.086~0.215,which improves the curve sensitivity while ensuring the focus range;under the influence of noise.This algorithm reduces the number of local extreme points that appear in traditional methods effectively,which maintains good stability and improves the defect of the traditional clarity evaluation method that is weak against noise.