A Vision System for Recognizing Objects in Complex Real Images
A new system for object recognition in complex natural images is here proposed. The proposed system is based on two modules: image segmentation and region categorization. Original images g(x,y) are first regularized by using a self-adaptive implementation
- PDF / 1,453,149 Bytes
- 11 Pages / 430 x 660 pts Page_size
- 92 Downloads / 195 Views
2
SISSA, Via Beirut 2-4, 34014 Trieste, Italy ICTP Programme for Training and Research in Italian Laboratories, International Center for Theoretical Phyiscs, Strada Costiera 11, 34014 Trieste, Italy {daliri,vanzella,torre}@sissa.it
Abstract. A new system for object recognition in complex natural images is here proposed. The proposed system is based on two modules: image segmentation and region categorization. Original images g(x,y) are first regularized by using a self-adaptive implementation of the Mumford-Shah functional so that the two parameters α and γ controlling the smoothness and fidelity, automatically adapt to the local scale and contrast. From the regularized image u(x,y), a piece-wise constant image sN(x,y) representing a segmentation of the original image g(x,y) is obtained. The obtained segmentation is a collection of different regions or silhouettes which must be categorized. Categorization is based on the detection of perceptual landmarks, which are scale invariant. These landmarks and the parts between them are transformed into a symbolic representation. Shapes are mapped into symbol sequences and a database of shapes in mapped into a set of symbol sequences. Categorization is obtained by using support vector machines. The Kimia silhouettes database is used for training and complex natural images from Martin database and collection of images extracted from the web are used for testing the proposed system. The proposed system is able to recognize correctly birds, mammals and fish in several of these cluttered images.
1 Introduction A major goal of computer vision is to make machines able to recognize and categorize objects as humans are able to do very quickly [1] even in very cluttered and complex images. There are several reasons that make this problem so difficult. The first reason is related to the uncertainty about the level of categorization in which recognition should be done. Based on the research made by cognitive scientists [2], there are several levels at which categorization is performed. Another reason is the natural variability within various classes. Moreover, the characterization should be invariant to rotation, scale, translation and to certain deformations. Objects have several properties that can be used for recognition, like shape, color, texture, brightness. Each of these cues can be used for classifying objects. Biederman [3] suggested that edge-based representations mediate real-time object recognition. In his view, surface characteristics such as color and texture can be used for defining edges and can provide cues for visual search, but they play only a secondary role in G. Bebis et al. (Eds.): ISVC 2007, Part II, LNCS 4842, pp. 234–244, 2007. © Springer-Verlag Berlin Heidelberg 2007
A Vision System for Recognizing Objects in Complex Real Images
235
the real-time recognition. There are two major approaches for shape-based object recognition: 1) boundary-based, that uses contour information [4], [5], [6], [7], [8], and 2) holistic-based representation, requiring more
Data Loading...
