영상 분할과 SIFT를 이용한 다중 객체 검출

Abstract

In this paper, a novel method for detecting multi-object using image segmentation and SIFT descriptor. Our research consists of four steps. In the first step, an image is segmented into several regions by JSEG algorithm. In the second step, for segmented regions, dominant colors and the corresponding color histogram are constructed. In the third step, by using dominant color and color histogram, candidate regions where objects may exist are identified. In the final step, real object regions are detected from candidate regions by SIFT matching. As experimental results, the proposed method shows the improved performance over exist ones.