Neuronic Algorithm for Recognition of Two-dimensional Binary Images in Noise Conditions
Recently, the great attention has been paid to the works aimed at the development of intelligent 2-D binary image recognition systems. The reason for that are the broad capabilities of computer vision systems, which now include some functions previously attributed to human only. As a rule, such recognition systems perform the optical and digital transformation of an image and its following recognition by using various algorithms. In many cases it is sufficient to analyze its contour image to achieve the effective recognition. In doing so, the internal elements of the object contour and background can be excluded as representing little information. Furthermore, in this case memory volumes decrease which are necessary for image recognition. At present the most widespread approaches to detect the object boundaries are: the statistic method, detection of contour curvature, and neural network theory application. However, for image description the first two methods require the determination of energy function using sufficiently large point neighborhoods, which results in a significant increase of computational costs. At the same time, the keen interest is displayed in the neural network theory, which is increasingly used to solve a number of engineering problems. This can be explained by the neural network capability of solving hardly formalizable problems, such as image recognition, and by their relatively simple practical realization. The work objective is to substantiate the neural network applicability for the image contour detection in noise conditions and compare between the results of neural and digital recognitions. Presented in this work the results of simulating the algorithm of object boundary detection on images testify the relevancy of neural network application to such problems. The neural network provides more quality solution for object boundary detection problems and enables to reduce the noise level.
Neuronic algorithm Image recognition Two-dimensional binary images Noise conditions
Ragim Mammadov Nizami Jafarov Matanat Hasanguliyeva
Department of Information Measurement and Computer Engineering, Azerbaijan State Oil Academy, Azerbaijan Azadliq 20 ave, Azerbaijan State Oil Academy, Baku AZ 1010
国际会议
沈阳
英文
491-493
2009-08-27（万方平台首次上网日期，不代表论文的发表时间）