Affinity Fusion Graph-Based Framework for Natural Image Segmentation

Abstract

This paper proposes an affinity fusion graph framework to effectively connect different graphs with highly discriminating power and nonlinearity for natural image segmentation. The proposed framework combines adjacency-graphs and kernel spectral clustering based graphs (KSC-graphs) according to a new definition named affinity nodes of multi-scale superpixels. These affinity nodes are selected based on a better affiliation of superpixels, namely subspace-preserving representation which is generated by sparse subspace clustering based on subspace pursuit. Then a KSC-graph is built via a novel kernel spectral clustering to explore the nonlinear relationships among these affinity nodes. Moreover, an adjacency-graph at each scale is constructed, which is further used to update the proposed KSCgraph at affinity nodes. The fusion graph is built across different scales, and it is partitioned to obtain final segmentation result. Experimental results on the Berkeley segmentation dataset and Microsoft Research Cambridge dataset show the superiority of our framework in comparison with the state-of-the-art methods. The code is available at https://github.com/Yangzhangcst/AF-graph.

An Improved Superpixel-based Fuzzy C-Means Method for Complex Picture Segmentation Tasks

“…In all, firstly bilateral filtering is used to implement information in the neighborhood. Then fuse the original feature with the filtered picture by (9) to ensure the feature in the original picture isn't lost. Next the SLIC is used to do a presegmentation which is shown in Fig.…”

“…In FCM_SCIM, an adaptive weight strategy is proposed which the weight is determined according to the filtered pixel value and original pixel value. Our method uses the strategy in [16] and the way to fuse is (9).…”

“…General segmentation aims to divide a picture into several parts according to its color, texture, intensity and so on. There are many method based on different theories for this task such as segmentation based on threshold value [6], region-based image segmentation [7], edge-based image segmentation [8], image segmentation based on graph theory [9] and so on. More and more attentions is paid for fuzzy c-means [10] on image segmentation, as the distribution of a picture's histogram meets the definition of it.…”

RNN-combined graph convolutional network with multi-feature fusion for tuberculosis cavity segmentation

“…Zhang et al [8] presented an affinity fusion graph framework that combines adjacency graphs and graphs based on nuclear spectral clustering for natural image segmentation. Lu et al [9] adopted an adjacency matrix of a Gaussian kernel function for image semantic segmentation.…”

“…In recent years, with the deepening of graph neural network research, graph neural network has also been successfully applied in the field of image segmentation. Examples include instance segmentation [7], semantic segmentation [8] and natural image segmentation [9]. Some researchers apply RNN to the field of image segmentation [10,11], making the segmentation task flexible and efficient.…”