【摘要】 提出了一种有效的模糊边界检测算法。首先,用滤波器组和改进的K-Means算法快速提取了图像的纹理基元特征;然后,采用模糊手段将图像的局部灰度信息、纹理信息和空间信息有机地融合起来,定义了一个边缘检测函数,求出每个像素所对应的模糊梯度值,并由此构成模糊梯度特征向量;最后,结合分类器的学习能力进行边界检测。实验结果表明,对于纹理和灰度边界混合的自然图像的边界检测问题,该算法是一种实用和有效的方法。更多还原

【Abstract】 This paper presents an effective algorithm for fuzzy boundary detection.First,texton feature is extracted using the filter bank and improved K-means clustering algorithm.Then this new algorithm incorporates local intensity information,texture information and spatial information in a fuzzy way to define a novel edge detection function.The fuzzy gradient value of each image pixel is computed from the edge-detection function,and furthermore,these values constitute fuzzy gradient feature vectors.And then a classifier is used to detect image boundary.Experiments performed on natural images with fuzzy intensity and texture show that the proposed algorithm is practical and effective.更多还原