Abstract: Edge detection is a hard problem in image processing and pattern recognition field because most methods suffer from noises. In order to provide a reliable edge extraction method, this paper proposes a novel approach. Firstly, the original image is enhanced by using a fuzzy logical function and histogram-enhancement like method, globally. Then, a local non-linear transform is carried out according a Gaussian-like function where two parameters are needed to be optimized. Finally, the transformed image is binaries and the edges are extracted. The paper also discusses the approach to optimize the parameters used by Gaussian-like function which can facilitate the edge extraction results. The proposed method can be used for image processing system and pattern recognition system.

Key words: fuzzy-logical, image enhancement, Gaussian function

摘要： 对图像边缘提取的问题进行了研究，针对以往边缘提取算法容易造成边缘断裂，提取较多噪声的缺点，提出一种利用模糊逻辑函数在全局上对图像进行增强，在局部进行边缘提取的算法。定义一个模糊逻辑函数，利用类似直方图均衡化的方法，对变换后的模糊逻辑图像进行对比度增强操作；在增强的图像上，对模糊逻辑函数进行局部非线性投影变换，并在变换基础上进行边缘图像的阈值检测，从而得到最终的边缘图像。在进行非线性变换的同时，利用梯度迭代法对非线性参数进行优化，从而保证获得最优的高斯模型，并提取出边缘信息。该方法在自动获得最优的系统参数同时，能准确有效地提取出图像的边缘，克服噪声对系统的干扰；并能广泛使用在基于边缘信息的图像处理系统和模式识别系统中。

关键词: 模糊逻辑, 图像增强, 高斯模型