关键词: 超像素分割, 聚类, K-means, K近邻

Abstract:

Superpixels algorithms have been applied to various fields of computer vision in recent years. Superpixels can capture the redundant information of the image and reduce the complexity of post-processing. As the preprocessing of images, superpixels segmentation needs to meet the demands of real-time properties and accuracy of image processing. We propose an improved fast SLIC segmentation algorithm in the framework of the SLIC algorithm to improve the efficiency of superpixels segmentation. We extract a small number of pixels of the original image to generate a small scale image by reducing the scale of the original image, and employ the SLIC algorithm to do superpixels segmentation of the downscaled image. After the primary superpixel segmentation, we perform K nearest neighbor classification on the pixels of the original image according to the segmentation results of the downscaled image and realize final superpixels segmentation. Experimental results show that with similar accuracy, the proposed algorithm is much faster than the SLIC algorithm when processing the same object.

Key words: