基于模糊神经网络的红外图像边缘提取算法

针对现有图像边缘提取算法存在的噪声平滑能力与边缘精确定位之间的矛盾,以及红外图像自身信噪比低、视觉效果模糊和对比度差等缺陷,利用模糊神经网络的学习、自适应和模糊处理等优点,提出了一种基于模糊神经网络的红外图像边缘提取方法。计算各像素点8个方向的基本梯度、左关联梯度和右关联梯度,并将其组成梯度数组,把8个方向的梯度数组作为模糊神经网络的输入信号,通过学习和模糊处理最终可获得相对精确的红外图像边缘。实验结果表明,该方法抗噪能力强,边缘保留完整且为单像素宽,在处理红外图像边缘提取上要优于其他算法。

Infrared image edge extraction based on fuzzy neural network

With regard to the contradiction between noise-smoothing capabilities and accurate positioning of image edge,which results from the existing image edge detection algorithms,a new infrared image edge detection method based on fuzzy neural network is proposed utilizing the fuzzy neural network′s learning,adaptive and fuzzy processing.We calculate basic gradient,left associated gradient and right associated gradient of each pixel in eight directions,which constitutes gradient array.The gradient arrays in eight directions are then set as input of fuzzy neural network.Finally the relatively precise infrared image edge is obtained by learning and obfuscation.Experimental results show that this method has powerful denoising capability.It can retain intact edge whose edge width is single-pixel level.That indicates the method is superior to other algorithms in dealing with infrared image edge.