Region-Based Saliency Detection and Its Application in Object Recognition

Abstract

The objective of this paper is twofold. First, we introduce an effective region-based solution for saliency detection. Then we apply the achieved saliency map to better encode the image features for solving object recognition task.To find the perceptually and semantically meaningful salient regions, we extract superpixels based on an adaptive mean shift algorithm as the basic elements for saliency detection. The saliency of each superpixel is measured by using its spatial compactness which is calculated according to the results of Gaussian Mixture Model (GMM) clustering. To propagate saliency between similar clusters, we adopt a modified PageRank algorithm to refine the saliency map. Our method not only improves saliency detection through large salient region detection and noise tolerance in messy background, but also generates saliency maps with welldefined object shape. Experimental results demonstrate the effectiveness of our method.Since the objects usually correspond to salient regions, and these regions usually play more important roles for object recognition than background, we apply our achieved saliency map for object recognition by incorporating saliency map into Sparse coding based Spatial Pyramid Matching (ScSPM) image representation. To learn a more discriminative codebook and better encode the features corresponding to the patches of the objects, we propose a weighted sparse coding for feature coding. Moreover, we also propose a saliency weighted max pooling to further emphasize the importance of those salient regions in feature pooling module. Experimental results on several datasets illustrate that our Weighted Sparse Coding based Spatial Pyramid Matching framework greatly outperforms Sparse Coding based Spatial Pyramid Matching framework, and achieves excellent performance for object recognition.

A Survey on Various Saliency Detection Methods

“…Finally, combine the local and global saliencies, and utilize the composition information to obtain the final saliency. In [9] Zhixiang Ren, Shenghua Gao, Liang-Tien Chia, and Ivor Wai-Hung Tsang presents a region-based solution for saliency detection applicable to better encode the image features for solving object recognition task. First use the adaptive mean shift algorithm to extract superpixels from the input image.…”

Discovering salient objects from videos using spatiotemporal salient region detection

“…Moreover, the formulation of color distribution cue for saliency is difficult these two approaches. Ren et al [2] clustered superpixels by Gaussian Mixture Model for saliency detection. In their work, saliency of a cluster is defined as its compactness which is estimated as the inter-cluster distance between clusters.…”

“…The existing saliency detection approaches work on either patch level [5,10] or region level [2,9,14,15] image abstractions. Each of these abstractions has their advantages.…”

“…Many computational visual attention models have been proposed over the years, to resemble the mechanism of the Human Visual System's remarkable ability to fixate conspicuous/salient regions from a visual scene. The field has gained considerable attention in the recent years, and has become an active area of research in computer vision due to its applications in object detection [1], object recognition [2], adaptive image and video compression [3], image retargeting [4,5], summarization of photo collections [4] and video summarization [6], where saliency detection is considered as an essential step towards achieving a vision goal. 3 temporal saliency detection [28,29,30].…”

Salient Region Detection Using Patch Level and Region Level Image Abstractions

“…Most existing salient region detection approaches operate on either patch level [1], [5] or region level [3], [6], [7] image abstractions. A major problem with patch level approaches is that, they often fail to suppress saliencies of textured background noises, and to highlight the saliency of objects uniformly.…”