Rooted Spanning Superpixels
- PDF / 4,942,343 Bytes
- 17 Pages / 595.276 x 790.866 pts Page_size
- 89 Downloads / 172 Views
Rooted Spanning Superpixels Dengfeng Chai1 Received: 25 April 2018 / Accepted: 9 June 2020 © The Author(s) 2020
Abstract This paper proposes a new approach for superpixel segmentation. It is formulated as finding a rooted spanning forest of a graph with respect to some roots and a path-cost function. The underlying graph represents an image, the roots serve as seeds for segmentation, each pixel is connected to one seed via a path, the path-cost function measures both the color similarity and spatial closeness between two pixels via a path, and each tree in the spanning forest represents one superpixel. Originating from the evenly distributed seeds, the superpixels are guided by a path-cost function to grow uniformly and adaptively, the pixel-by-pixel growing continues until they cover the whole image. The number of superpixels is controlled by the number of seeds. The connectivity is maintained by region growing. Good performances are assured by connecting each pixel to the similar seed, which are dominated by the path-cost function. It is evaluated by both the superpixel benchmark and supervoxel benchmark. Its performance is ranked as the second among top performing state-of-the-art methods. Moreover, it is much faster than the other superpixel and supervoxel methods. Keywords Superpixels · Segmentation · Spanning forest
1 Introduction Superpixels have become effective alternative to pixels in the past decade. They result from image oversegmentation, which is dedicated to reducing image complexity while avoiding undersegmentation (Ren and Malik 2003). An image is oversegmented into many perceptually meaningful segments such that each segment covers a local region consisting of some connected similar pixels, and each segment is called as a superpixel. Superpixels have two prime advantages over pixels. One advantage is the perceptual meaning. In contrast with raw pixels generated by digital sampling, superpixels are formed by pixel grouping, Communicated by Yuri Boykov. This work was supported by the National Natural Science Foundation of China (No. 41571335). The source code for RSS algorithm can be found at https://github. com/dfchai/Rooted-Spanning-Superpixels.
B 1
Dengfeng Chai [email protected] Key Laboratory of Geoscience Big Data and Deep Resource of Zhejiang Province, School of Earth Sciences, Zhejiang University, No. 38, Zheda Road, Hangzhou 310027, Zhejiang, China
whose principles are based on the classical Gestalt theory (Wertheimer 1938) assuring superpixels enhanced perceptual meaning. This characteristic facilitates defining higher order potentials, high order conditional random fields and associative hierarchical random fields (Arnab et al. 2016). The other advantage is the complexity. Since many pixels are grouped into one superpixel, the number of superpixels is much smaller than that of pixels. When superpixels instead of pixels serve as atoms, the size of an image is reduced greatly. The size reduction can accelerate the processing in subsequent tasks, and in turn, it is possible to employ so
Data Loading...
