二值图像的自适应窗口边缘检测技术

英国Aleksander等人用RAM网络实现的自适应模式识别系统可以认为是一种逻辑神经网络,本文详细描述了利用这种模型实现的自适应窗口边缘2检测技术,包括模型的理论分析,边缘检测种种的各种实现方法和实验结果。基于这种模型的边缘检测技术的基本思想是:首先对窗口边缘检测技术,包括模型的理论分析,边缘检测的各种实现方法和实验结果。训练一个或多个对目标物体边缘敏感的模式,然后对图像逐点扫描和测试,根据窗口对特定扫描区域的响应实现对边缘的检测。其过程和模式识别中的训练—识别过程相似。文中讨论了“中心块”和“边缘训练”窗口的响应行为以及窗口大小的选择、响应阀值的确定和N元的选用等具体方法。并给出了相应的实验结果。研究表明,这种方法对边缘检测和噪音滤除都具有较好的效果,且硬件容易并行实现。

Adaptive Window Approaches to the Edge Detection on the Binary Images

The adaptive pattern recognition system using RAM array, implemented by I. Aleksander et al. ,can be considered as a logical neural network. The paper has described the adaptive window edge detection techniques based on this model,including the theoretical analysis,a variety of available approches to the edge detection and experimental results. The basic ideas of these techniques are first to train one or several patterns sensitive to edges in images,then to scan and test the specified image pix-wise for dectecting edges in accorance with the response of the window to the specified area in the image,the procedure is similar to the one of trining-and-test in pattern recognition. The response behaviour of both "centre-on" and "edge-training" windows with the selection of the N-tuple,the choice of the window size and the determination of the response threshold has been discussed and related experimental results given as well. It has been shown that these techniques have better effects specially on the edge detection of binary images with noise and are easy to implement in parallel with hardware.