Loading...
Please wait, while we are loading the content...
Similar Documents
Chapter 2 Range Image Segmentation Using a Relaxation Oscillator Network 2.1 Introduction 2.2 Overview of the Legion Dynamics 2.2.1 Single Oscillator Model
| Content Provider | Semantic Scholar |
|---|---|
| Abstract | In this chapter, a locally excitatory globally inhibitory oscillator network (LEGION) is constructed and applied to range image segmentation, where each oscillator has excitatory lateral connections to the oscillators in its local neighborhood as well as a connection with a global inhibitor. A feature vector, consisting of depth, surface normal, and mean and Gaussian curvatures, is associated with each oscillator and is estimated from local windows at its corresponding pixel location. A context-sensitive method is applied in order to obtain more reliable and accurate estimations. The lateral connection between two oscillators is established based on a similarity measure of their feature vectors. The emergent behavior of the LEGION network gives rise to segmentation. Due to the flexible representation through phases, our method needs no assumption about the underlying structures in image data and no prior knowledge regarding the number of regions. More importantly, the network is guaranteed to converge rapidly under general conditions. These unique properties lead to a real-time approach for range image segmentation in machine perception. The results presented in this chapter appeared in [83]. Image segmentation has long been considered in machine vision as one of the fundamental tasks. Range image segmentation is especially important because depth is one of the most widely used cues in visual perception. Due to its practical importance, many techniques have been proposed for range image segmentation, and they can be roughly classified into four categories: 1) edge-based algorithms 4) global optimization of a function [73]. Edge-based algorithms first identify the edge points that signify surface discon-tinuity using certain edge detectors, and then try to link the extracted edge points together to form surface boundaries. For example, Wani and Batchelor [136] introduced specialized edge masks for different types of discontinuity. Because critical points, such as junctions and corners, could be degraded greatly by edge detectors, they are extracted in an additional stage. Then surface boundaries are formed by growing from the critical points. As we can see, many application-specific heuris-tics must be incorporated in order to design good edge detectors and overcome the ambiguities inherent in linking. Region-based algorithms were essentially similar to region-growing and split-and-merge techniques for intensity images [152], but with more complicated criteria to incorporate surface normal and curvatures which are critical for range image segmen-tation. A commonly used method is iterative surface fitting [4][71][51]. Pixels are first coarsely classified based on the sign of mean and Gaussian surface … |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://fsvision.fsu.edu/publications/papers/dissertation/chapter02.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Chapter |
