Enhancing Edge Detection of Depth Image by Bilateral Filter and Morphological Operations Thai Leang Sung Department of Computer Science and Engineering Chonbuk National University 567 Baekje-daero, Deokjin-gu, Jeonju-si, Jeollabuk-do, South Korea thaileang@jbnu.ac.kr Hyo Jong Lee Department of Computer Science and Engineering, CAIIT Chonbuk National University 567 Baekje-daero, Deokjin-gu, Jeonju-si, Jeollabuk-do, South Korea hlee@chonbuk.ac.kr Abstract—Edge detection in depth image remains a challenge in computer vision. In this paper, we propose an enhancement of depth edge detection using bilateral filtering and morphological operations such as erosion and dilation. The edge detection is done based on Canny Edge detection principle. The results have shown that this method provided better results than the method without the enhancement. Keywords—kinect, depth image, edge detection I. INTRODUCTION Edge detection in RGB image has been successfully researched, but it is still a challenge for depth edge detection in computer vision field. Accurate edge detection from a depth image is essential for some object detection processes [1], which are dependent on a model of a particular shape. A proper edge detection process can be used for various Human action analysis problems in a real environment such as walking, spotting and sitting [2]. However, existing edge detection process in depth images cannot be applied in these types of situations due to some limitations. Some methods of edge detection in depth image failed to deliver noise-free depth images; thus, proper edge detection cannot be achieved. In this paper, we proposed a method that can detect edges from depth images much better. We use Canny edge detection [3] to detect continuous edges along with the incorporation of morphological operation [4]. This operation generally consists of two operators; erosion and dilation. The first operation denoted as opening, smooths the contour object, break narrow strips and eliminates thin protrusions. The second operation, called closing, also smooths contours but in contrast with opening; it fuses thin discontinuities, eradicate trivial holes and fills gaps in the contour. In section 2, we will explain about our depth detection algorithms i.e. smoothing algorithm, morphological operation, and canny edge detection and its modification. Section 3 is our experimental results for different scenes and compare the results 1 Middlebury and NYUDepthV2 with the method that doesn’t have the enhancement. Finally, Section 6 concludes this research paper. II. PROPOSED APPROACH In Figure 1, we have illustrated the process of the proposed system. We have input depth image 1 input to the detection process. Then we obtain an output edge image as a result. In the detection processing, the depth image will go through smoothing algorithm before the process of edge detection and morphological operation are applied. Figure 1 Proposed Approach In the following we will describe the depth edge detection process step by step respectively. A. Smoothing Algorithm Smoothing algorithm is used to reduce the level of noises in depth image. Here, we use Bilateral Filtering [5]. (a) (b) Figure 2. (a) Original Depth Image, (b) Filtered Image