Abstract: To address the issues of flat curves, multi peaks, and misjudgments in traditional sharpness evaluation methods, a gradient threshold based sharpness evaluation method combining multiple region windows is proposed. The method utilizes the contrast and weighted variance of the gradient image to calculate a segmentation threshold and extract edge points. An improved energy gradient function is then applied to evaluate the sharpness of the region image at the edge points. Different weights are assigned to the evaluation values of various regions to obtain the final evaluation result for the entire image. Comparing this method with traditional approaches, the results show that when using the same evaluation function, the multi region window produces a unimodal evaluation curve without misjudgments, in contrast to the traditional window. The sensitivity factor of this method is, on average, 55.6% higher, and the steepness is 58.3% higher than the optimal results from traditional methods. Therefore, the proposed method can provide an effective reference for automatic focusing applications in space optical sensors.

Key words: autofocus, image processing, sharpness evaluation, focusing window