Specialized cellular structures for image contour analysis
- PDF / 533,690 Bytes
- 10 Pages / 595.276 x 793.701 pts Page_size
- 97 Downloads / 177 Views
SPECIALIZED CELLULAR STRUCTURES FOR IMAGE CONTOUR ANALYSIS UDC 658.012:681.32
S. N. Belan
Abstract. This article considers principles of construction of a multichannel technical vision system using cellular automata. Based on the proposed concept of a geometric type, images invariant to rotations, scaling, and dynamic changes are recognized. Methods based on cellular technologies are proposed for the construction of geometric-type images. Keywords: image, recognition, geometric type, invariance, orientation, scale, cellular automaton. INTRODUCTION The present state of the art of technical vision is characterized by many approaches that are basically focused on the realization of functions and structures modelling the human visual canal of a sort [1–5]. Much attention is given to artificial neural networks (ANNs) that consist of neural-like elements and connections with changing weights between them [6]. Many neuroarchitectures and various paradigms and training algorithms are well known. However, in solving complicated problems of processing visual information, supporters of ANNs encounter some obstacles and problems that, first of all, lie in a bounded collection of functions required for special data processing and also in time- and labor-consuming training algorithms. It is significant that, in the field of image processing and recognition, despite the existing variety of methods and means, they are not interchangeable (for example, systems of biometric identification of persons from fingerprints and irises and face identification cannot distinguish between symbols and various geometric figures and also cannot replace one another). At the same time, a modern computer cannot distinguish between a cat and a dog [7] and, therefore, more and more specialists are convinced of the necessity of searching for new architectural and hardware approaches together with software ones that would promote the real-time processing of video data. In constructing image processing systems, the need arises to solve some specialized problems of efficient preliminary data preparation. Among them, efficient video data compression and achievement of high speed are main problems. Some of the most efficient means of processing video data are cellular automata (CAs). Since a CA possesses the homogeneity property and can realize any connections between cells of its environment, the possibility of both image contouring and image processing depending on brightness parameters and color characteristics is provided by reconfiguring it during data processing. This article is devoted to the solution of the problem of realization of preprocessing and efficient data preparation on specialized cellular structures. They realize basic operations of image preprocessing and form data representations making it possible to analyze images invariant with respect to affine scale and rotation transformations. PRINCIPLES OF CONSTRUCTION OF A TECHNICAL VISION SYSTEM In a technical vision system (TVS), the following functions of neurophysiological systems are ta
Data Loading...
