基于联合约束策略和稀疏表示的图像分割

基于像素级的交互式图像分割算法对初始种子位置和噪声敏感,同时仅基于超像素的分割方法无法保留图像细节经常导致分割结果出现欠分割问题。针对上述问题,提出超像素/像素约束和稀疏表示的图像分割模型。该方法利用高斯函数分别对像素和超像素构造了相互约束的代价函数,引入了稀疏分解对模型进行优化以提升模型对图像噪声的鲁棒性,最后利用联合优化策略对代价函数求解估计出目标和背景标记实现目标提取。实验结果表明,与现有的分割方法相比,提出的方法能获得较好的分割效果,对高斯噪声和椒盐噪声具有较强的鲁棒性。

Image segmentation based on joint constraint strategy and sparse representation

The interactive image segmentation algorithm based on pixel-level is sensitive to initial seed position and noise.However,the image segmentation method only oriented super-pixels is unable to retain image details so that under-segmentation could produce.Aiming at above problems,this paper proposed the segmentation model based on superpixel/pixel constraint and sparse representation.The method firstly defined two mutual constraint cost functions utilized the Gaussian model for super-pixels and pixels respectively.Meanwhile,it introduced constraint condition oriented on sparse decomposition to optimize the proposed model to avoid over-segmentation.Finally,this paper utilized joint optimization method to obtain the solution of the proposed model and estimated the object and background labels to realize object extraction.Experiment results de-monstrate that the proposed method can obtain better segmentation results compared with the state-of-the-art methods and is more robust to noise.