关键词: 主动轮廓模型, 显著性检测, 图像分割, 高斯混合模型

Abstract:

The traditional active contour method cannot highlight the saliency of the segmented region. At the same time, the target in the saliency map obtained by the saliency detection algorithm has a higher SNR. This paper proposes an active contour image segmentation combining saliency. First, superpixels are obtained by linear spectral clustering segmentation. Superpixels are used as processing units to obtain better saliency maps based on a graph theory-based manifold ranking algorithm. Then, the Gaussian mixture model is introduced into the curve evolution process of the active contour, and the average gray value inside and outside the curve is calculated. Thus, a new active contour energy equation is obtained through the Gaussian mixture model and saliency information, and the level set method is used to guide the segmentation. The final segmentation result is obtained. Experimental results show that the image segmentation method proposed in this paper can segment images quickly and efficiently.

Key words: active contour model, saliency detection, image segmentation, Gaussian mixture model