一种基于加权切比雪夫距离的图像分割方法

An Image Segmentation Method Based on Weighted Chebyshev Distance

传统的图像分割相似性度量方法,存在片面考虑图像像素值的大小,或仅考虑像素点间的距离等局限性。同时大多数图像,由于光线角度、背景干扰等原因,需用精度更高的计算方法去除噪声点。为提高图像分割算法的性能,综合考虑图像像素点的像素值大小和像素点间的距离,提出了一种基于加权切比雪夫距离的图像分割算法(Image Segmentation based on Weighted Chebyshev Distance,ISWCD).首先将图像梯度信息和邻域信息结合使用阈值法去除噪声,从对应图像的RGB空间像素矩阵中提取特征向量,然后计算加权切比雪夫距离,从而得出相似矩阵,最终利用谱聚类实现图像分割。在BSDS300和VOC2012两个数据集上进行实验,结果表明ISWCD算法性能优于传统图像分割算法。

The traditional image segmentation similarity measurement method has some limitations such as considering the size of the image pixel value unilaterally,or only considering the distance between pixels.At the same time,due to the angle of light,background interference,etc.,most images require a more accurate calculation method to remove noise points.In order to improve the performance of the image segmentation,considering the value of the image pixels and the distance between the pixels,this paper proposes an image segmentation algorithm based on weighted Chebyshev distance(ISWCD).Firstly,the image gradient information and the neighborhood information are combined to remove noise using the threshold method.The feature vectors are extracted from RGB space pixel matrix of the corresponding image,and then the weighted Chebyshev distance is calculated to obtain a similarity matrix.Finally,spectral clustering is used to realize the image segmentation.Experiments show that the performance of our ISWCD algorithm is superior to traditional image segmentation algorithms on two datasets BSDS300 and VOC2012.