鉴于传统的灰色边缘检测算法在阈值选取上需要人工干预, 不具备自适应能力, 提出了基于灰色简化 B型关联度的自适应阈值选取进行图像边缘检测新算法。首先, 用标准差化矩阵算子对图像灰度值进行预处理; 其次, 在 3×3的像素模板上, 将预处理后的中心像素点及周围 8个像素点的数值一维化作为比较序列, 并将这 9个像素点的均值作为参考序列; 最后, 利用简化 B型关联度计算两者之间的灰关联度, 根据迭代算法求取灰色简化 B型关联度的最佳阈值来检测图像边缘。实验结果表明, 所提算法对灰度变化剧烈的图像具有较强的适应性, 检测边缘清晰准确, 比传统的邓氏相关度边缘检测算法能够更好的抑制噪声。

Because the traditional gray edge detection algorithm on the threshold selection needs manual intervention and does not have the adaptive ability, a novel algorithm is proposed to adapt threshold to detect edges on gray images based on the gray correlation degree of simplified B-mode. At first, the gray-scale value of image is preprocessed by using the standard deviation of matrix operator. Secondly, in the 3×3 pixels template, the pretreatment gray-scale values of nine nearby pixels is converted to one dimensional data which acts as comparative sequence. Moreover, the mean of those nine pixels acts as referenced sequence. At last, the gray correlation degree of simplified B-mode of incidence between them was calculated, which is the basis of image edge detection. For images with great gray variation, the experimental results show that the proposed algorithm not only has strong adaptability, is able to clearly and accurately detect the edge, but also can effectively restrain noise more than several classical Deng interrelatedness edge detection algorithms.