Improved Simple Linear Iterative Clustering Algorithm Using Hsl Color Space
Fan Su,Hui Xu,Guodong Chen,Zhenhua Wang,Lining Sun,Zheng Wang
DOI: https://doi.org/10.1007/978-3-030-27541-9_34
2019-01-01
Abstract:Image processing is a very important technical support in robotic vision. As a preprocessing step for image processing, superpixel segmentation is one of the significant branches of image segmentation. Simple linear iterative clustering (SLIC) algorithm, as a widely used superpixel segmentation algorithm, can help to deal with boundary adherence and reduce computational cost for image segmentation. However, the segmentation result of the original SLIC algorithm fails to adhere well to boundary of object, causing undersegmentation sometimes. Therefore, the work proposed an improved method based on SLIC to address the problem of undersegmentation. First, HSL color space was introduced to have a better recognition and processing of color instead of CIELAB color space. In addition, adding flexible combinations of weight coefficient for HSL can achieve different results. Finally, an edge detection strategy was used to enhance the accuracy of superpixel segmentation. The quantification effect of the proposed method was verified using the dataset BSDS500. The experimental results show that the improved algorithm has better accuracy and efficiency.