An investigation on the stability of object extraction

The stability of object extraction is put forward and analyzed, It is pointed out that one of the reasons why the object extracting algorithm based on the statistical decision making theory may be unstable is lack of the discriminating ability of the time-space variable object-background condition and self-adjustment of the appropriate region to be segmented. Usually the algorithm is based on the absolute image grey level or the absolute grey level variation and its statistical character. According to the principle of vision, a self-adaptive object extracting model and adaptive clustering-based thresholding algorithm (ACBTA) with controllability for sequential images are proposed. Experimental results show that, compared with certain conventional methods, the model and algorithm proposed are better in that they are ideally stable for extracting objects in images and image sequences with a complicated background.