Shape extraction from gray-valued images by concept construction networks

本研究的目的是建立一个能从灰度图像中提取各种二维和三维目标的神经网络系统。本研究包含以下三个步骤：1.发展模式辨识方法，以从方块与白色影像中撷取二维与三维物件。在这一步中，相当简单的形状，如直线，圆和字符是considered.2.to开发的神经网络，可以提取复杂的2D和3D对象从灰度值的图像通过使用在第一步中获得的结果。例如，从单色photographs.3.to中提取人脸区域，建立了一个形状提取神经网络系统，可以从图像中提取出各种二维和三维物体，该研究可以应用于计算机视觉领域的图像识别和理解。

The purpose of this research is to make a neural network system which can extract various two dimensional and three dimensional objects from gray-valued images. This research consists of following three steps;1.to develop pattern recognition methods for extraction of 2D and 3D objects from block and white images. In this step, rather simple shapes such as straight lines, circles and characters are considered.2.to develop neural networks which can extract complex 2D and 3D objects from gray- valued images by the use of the result obtained in the first step. For example, face regions are extracted from monochromatic photographs.3.to make a shape extraction neural network system which can extract various kind of 2D and 3D objects from images.This research can be applied to recognition and understanding of images in the field of computer vision.