自适应梯度重建分水岭分割算法

Adaptive gradient reconstruction for watershed based image segmentation

目的针对灰度分水岭算法存在过分割且难以直接应用到彩色图像分割的问题,提出一种自适应梯度重建分水岭分割算法。方法该方法首先利用PCA技术对彩色图像降维,然后计算降维后的梯度图像,并采用自适应重建算法修正梯度图像,最后对优化后的梯度图像应用分水岭变换实现对彩色图像的正确分割。结果采用融合了颜色距离、均方差和区域信息的性能指标和分割区域数对分割效果进行评估,对不同类型的彩色图像进行分割实验,本文算法在正确分割图像的同时获得了较高的性能指标。与现有的分水岭分割算法相比,提出的方法能有效剔除图像中的伪极小值,减少图像中的极小值数目,从而解决了过分割问题,有效提升了分割效果。结论本文算法具有较好的适用性和较高的鲁棒性。

Objective A watershed segmentation algorithm based on adaptive gradient reconstruction is proposed to address over-segmentation in the gray-scale watershed algorithm and to simplify its application on color images.Method Principal component analysis is used to reduce the dimensionality of color images.The gradient of low-dimensional images is calculated.The gradient image is modified by applying an adaptive gradient reconstruction algorithm.Watershed transformation is employed to the optimized gradient image to achieve color image segmentation.Result Use performance indicators and the numbers of segmented regions that contain color distance,mean square error and region information to evaluate the segmented results.For different types of color images.This algorithm can correctly segment different types of color images.Compared with the existing watershed segmentation algorithms,the proposed method can effectively eliminate the pseudo minimum caused by irregular details and noise.It can also overcome over-segmentation,thereby improving segmentation accuracy.Conclusion The proposed algorithm has good applicability and robustness.