关键词: 梯度向量流, 噪声敏感性, 水平集, 轮廓曲线, 细节信息

Abstract:

Aiming at the problem that the current active contour model was sensitive to noise, and it was difficult to realize accurate segmentation for weak boundary image, we proposed a new active contour model based on gradient vector flow. Firstly, contourlet transform was used to denoise image, and interference of noise to image segmentation was solved. Secondly, an instruction function was introduced to active contour model, which was used to describe the relationship between the vector field and the contour curve, and image was segmented by contour curve evolution. Finally, experiments were carried out to verify the performance of the proposed image segmentation model. Experimental results show that the proposed method can achieve fast and accurate segmentation for many kinds  of images, and segmentation accuracy and anti noise ability are better than other active contour models.

Key words: gradient vector flow, noise sensitivity, level set, contour curve, detail information