抄録
Recently, there have been many brutal crimes. Therefore, surveillance cameras have been installed at various places. This has led to the volume of image records being very large. Therefore it is necessary to develop an automatic method for checking suspicious persons. It is difficult to extract parts containing people from an image sequence. A lot of methods for extracting people regions have been developed. However the conventional methods need many images or precise background images. In this paper, we propose a method to extract people regions from an image by applying the Hough Transform twice. In the first step, people regions are obtained by the first Hough Transform of direction cords. The first Hough Transform pertains to connecting of the outline of the people regions and extracts the pixels of the people regions. In the next step, people regions are decided by the second generalized Hough Transform which pertains to the shapes of heads and shoulders. Some experimental results that indicate the effectiveness of the proposed method are also reported.
