Automatic region segmentation based on intensity distribution analysis for image sequences

摘要

An initial region segmentation method based on the mixture Gaussian model is newly developed. First, distribution parameters that correspond to objects in the scene are estimated. The parameters are estimated from intensity histogram data using the EM algorithm. Next, range parameters for region segmentation are determined using the estimated distribution parameters. Finally, the images are segmented by these range parameters. Experimental results of real scenes using the PC program show the effectiveness of the method. This method is expected to be available for a variety of object recognition systems for various industrial applications, such as visual surveillance, the fixed ground sub-systems of the intelligent transport system (ITS), and welfare support systems.