基于Sobel算子的医学图像边缘检测研究

Research of edge detection algorithm for medical image based on Sobel

边缘检测是图像处理过程的关键技术。由于医学图像的特殊性,检测边缘的准确性对疾病的诊断和治疗有着重大的影响。针对传统Sobel算法存在定位不精确、提取边缘较粗等不足,提出了一种改进算法。算法在传统Sobel算子模板基础上增加了45°方向和135°方向两个模板,提高了边缘定位的精度,采取局部梯度均值作为阈值对初始梯度图像进行局部梯度筛选,局部弱边缘得到增强,然后对处理后梯度图像进行细化和提取,得到边缘图像。实验证明,算法获取的图像边缘与传统Sobel算法相比,具有定位准确、边连续性好和边缘较细等优点,在医学图像处理中具有一定的实用性。

Edge detection is a important technique in image processing. Due to the particularity of medical images, accuracy of edge detection has a significant impact on the diagnosis and treatment of the disease. In traditional Sobel operator, detection edges is not accuracy and thick, propose an improved algorithm. Algorithm increases 135° and 45° directions templates with new weights based on the original two directions templates, improve the edge localization accuracy. Use the local grade mean as threshold to filter gradient for initial gradient image, local weak edge is strengthened, than refine the edge of the filtered image and extract the image edge. Experiments show that the new algorithm for edge detection has good accuracy, clear contours and thinner line compared with the traditional Sobel algorithm, which has some practicability in medical image processing.