一种新的基于高斯偏移差分过滤器与方向图的图像分割算法

A New Image Segmentation Algorithm Based on Difference of Gaussian Offset Filter and Directed Graph

为了使图像分割算法通用性更强,利用高斯偏移差分过滤器模拟人类视觉机制模型寻找图像中有纹理区域,使用方向图算法界定目标边缘,最终得到图像分割结果.该算法能有效地克服图像背景上的噪声以及前景目标轮廓不明显的缺陷,成功分割提取出图像的前景目标.试验证明,该算法鲁棒性强,具有一定通用性且无需先验知识.

To make a more generic image segmentation algorithm, a new image segmentation algorithm is proposed, in which the difference of Gaussian offset filter is used to find the texture area of image by simulating the model of the human visual system, the directed graph is used to calculate the object edge, and the segmentation result can be obtained. The defects of image, such as background noise and unconspicuous object contour, can be effectively overcome. The experimental result shows that the presented algorithm is robust and universal, and the correct separation result can be obtained without any prior knowledge.