Auto-focus windows selection algorithm for optical microscope

Aiming at the problem that traditional auto-focus windows selection algorithms are susceptible to content distribution, impurities and noise, an auto-focus windows selection method depending on variation of content pixels is proposed. The method takes a weighted average of the amount of gray-leveldifference pixels and the amount of edge pixels as the number of content pixels, then measures image content of sub-block in blurred state and divides focusing windows by the variation of content pixels. In the meantime, the proposed method estimates content distribution information of original image using image content of each sub-block in downsampling image. Besides, the proposed method detects out-of-blurred areas using combined detection of local standard deviation and sharp edge pixels, and it effectively excludes the misjudgments of focal plane caused by the impurities on coverslips. Compared with the traditional auto-focus windows selection algorithms, the proposed method could obtain effective focusing windows for each microscopic sequence varying in richness and distribution of the image content, the corresponding mean gradient of pixels is higher, and the acquired focus measure curve has fewer local extreme values and sharp performance, therefore, the proposed method has more advantages in success ratio and robustness.