Due to distortion, noise, segmentation errors, overlap, and occlusion of objects in digital images, it is usually impossible to extract complete object contours or to segment the ...
Abstract. Human perception of shape is based on visual parts of objects to a point that a single, significant visual part is sufficient to recognize the whole object. For example,...
Abstract. We describe an efficient approach to construct shape models composed of contour parts with partially-supervised learning. The proposed approach can easily transfer parts ...
We propose a novel framework for contour based object
detection and recognition, which we formulate as a joint
contour fragment grouping and labeling problem. For a
given set of...
ChengEn Lu, Longin Jan Latecki, Nagesh Adluru, Hai...
In this paper, we propose a novel framework for contour based object detection. Compared to previous work, our contribution is three-fold. 1) A novel shape matching scheme suitabl...