Aimed at focusing precision problems under a strong shot noisy background image, the processing of focusing evaluation function is deeply studied. Based on the experiment of conventional image sequences, sum of modified laplacian (SML) and laplacian focusing evaluation functions have better performance. Aimed at strong shot noisy image sequence, an algorithm is provided that this image sequence is implemented with median filtering. Then this sequence is over-segmented using the watershed algorithm in order to obtain area blocks. The pixel value is substituted by the mean value of this area block, which can reduce the shot noise′s influence. Finally, this image sequence is processed by SML or laplacian focusing evaluation function. Experimental results show that this processing′s accuracy is close to the image sequence without strong shot noise. This algorithm has good robustness.