Enhanced Autofocus Algorithm Using Robust Focus Measure and Fuzzy Reasoning

A new passive autofocus algorithm consisting of a robust focus measure for object detection and fuzzy reasoning for target selection is presented. The proposed algorithm first detects objects distributed in the image using a mid-frequency discrete cosine transform focus measure and then selects the target object through fuzzy reasoning with three fuzzy membership functions. The proposed algorithm is designed as full digital blocks and fabricated using 0.35-mum CMOS technology. Experimental results obtained from various samples show that the present method is insensitive to Gaussian and impulsive noises and able to improve the quality of the image by focusing to the appropriate target object.