Computer Vision News - August 2016

Every month, Computer Vision News shows you a nice trick that will make your work easier. This time, our engineers have decided to tell you how to detect image edges with a combination of three techniques . Edge detection is a critical component of many vision applications (e.g. image segmentation, object detectors). Despite the abundant research that has been carried out on edge detection over the past 50 years, it still remains a fundamental task and a source for research in computer vision. The trick of this months will show you how to detect image edges with advanced techniques. The method we will demonstrate is a combination of three techniques: A. Fast Edge Detection Using Structured Forests. B. Edge Superpixels. C. The Ultrametric Contour Map (UCM). The source code for all those methods can be found in the Structured Edge Detection Toolbox V3.0 . Let’s take a toy image to play with. The image below with houses, sky (although not blue, but still sky), a green stretch of grass, flowers and trees. Using the well-known canny or sobel edge detectors would give us many non- useful edges: 12 Computer Vision News Trick Advanced Edge Detection Methods Trick