基于分区梯度的医学图像边缘检测算法

Edge detection algorithm based on partition gradient for medical image

在医学图像处理中,边缘检测的准确性直接影响到疾病的诊断和治疗。针对传统边缘检测算法存在的方向性不强及检测边缘较粗的问题,提出了一种分区梯度的医学图像边缘检测算法。算法将5×5检测窗口按照中轴线和对角线划分为8个区域(4对对称区域),每对区域对应一个方向模板,通过模板分别与窗口图像进行卷积运算获得0°,45°,90°和135°方向的方向梯度,取最大值作为窗口中心点的梯度值。对梯度图像采用了改进的非极大值抑制方法进行细化,最后采用阈值法提取图像边缘。实验结果表明,该算法检测的医学血液细胞图像边缘方向性较强,边缘较细,检测效果明显优于传统Sobel算法。

In medical image processing,the accuracy of edge detection affects the diagnosis and treatment of the disease di-rectly. Since the traditional edge detection algorithm has weak of direction and coarseness of detecting edge,the edge detection algorithm based on partition gradient for medical image is proposed. The proposed algorithm divides the 5×5 detection window in-to eight areas（four pairs of symmetric areas）according to the central axis and the diagonal,each pair of areas is corresponding to a direction template,the direction gradients of 0°,45°,90° and 135° are obtained by convolution operation of the templates with the window image respectively,it takes the maximum of these direction gradients as the gradient value of the window center point. The gradient image is refined by the improved non-maximum suppression method,the image edge is extracted by threshold method. Experimental results show that the image edge of medical blood cells has good direction which tested by the proposed al-gorithm,the detection effects are obviously superior to the traditional Sobel algorithm.