基于自动样本和PSO优化组合核的图像分割

Image Segmentation of Optimized Combined Kernel Based on Automatic Sample and PSO

彩色图像分割在图像处理中占据重要的位置。为避免手动选取图像样本的不可靠性,文中采用K-means预分类图像,再通过Matlab编程自动选取图像的HSV颜色空间的特征样本。文中提出分块的思想:对彩色图像处理前进行分块处理,可判断为背景或前景的子块直接输出,对需要分割的子块运用支持向量机(SVM)方法进行训练分割。线性组合全局核函数和局部核函数,选出适合图像分割的最优组合核函数并引入粒子群算法优化支持向量机(PSO-SVM)的核参数c、g。实验表明,文中方法是有效的,图像分割精度满意、稳定。

Color image segmentation occupies an important position in image processing.To avoid the unreliability of image samples with manual selection,use K-means for image＇s pre-classification,and then select the image＇s HSV color space features via Matlab programming automatically.Present the idea of the block： process the color image with partitioning firstly; then output the block images that can be judged as background or foreground directly; use Support Vector Machine（ SVM） method for training and segmenting the remaining block images.With the linearly combination of the global and local kernel,select the optimal combination kernel function for image segmentation.Introduce the Particle Swarm Optimization（ PSO） to optimize the parameters in combined kernel.The experimental results show that the proposed method is effective.The image segmentation accuracy is satisfactory and stable.