Efficient Salient Region Detection with Soft Image Abstraction

By Ming-Ming Cheng, Jonathan H. Warrell, Wen-Yan Lin, Shuai Zheng, Vibhav Vineet, Nigel Crook. IEEE International Conference on Computer Vision, 2013

595 citations recorded by the source corpus.

Detecting visually salient regions in images is one of the fundamental problems in computer vision. We propose a novel method to decompose an image into large scale perceptually homogeneous elements for efficient salient region detection, using a soft image abstraction representation. By considering both appearance similarity and spatial distribution of image pixels, the proposed representation abstracts out unnecessary image details, allowing the assignment of comparable saliency values across similar regions, and producing perceptually accurate salient region detection. We evaluate our salient region detection approach on the largest publicly available dataset with pixel accurate annotations. The experimental results show that the proposed method outperforms 18 alternate methods, reducing the mean absolute error by 25.2% compared to the previous best result, while being computationally more efficient.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/4d079f3e6cee42fbf8bf8722d27307d615a6273e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Computer Vision and Pattern Recognition

Citations in the corpus

DenseCut: Densely Connected CRFs for Realtime GrabCut (2015, methodology): “In the salient object detection community, more efﬁcient GMM estimation methods have recently been developed [CWL ∗ 13].”
Saliency detection by multi-context deep learning (2015, methodology): “Methods for comparison include DRFI [25], HS [56], GBMR [57], PCAS [41], CEOS [40], GC [12], and SF [44].”
Surface defect saliency of magnetic tile (2018, background): “…models are designed to predict human ﬁxations [12–14], driven by computer vision applications such as image segmentation [15], image abstraction [16], image manipulation [17], object retrieval [18], scene classiﬁcation [9], 3D model simpliﬁcation [19], or only focused at detecting the saliency…”
DeepSaliency: Multi-Task Deep Neural Network Model for Salient Object Detection (2015, methodology): “We qualitatively and quantitatively compare the proposed approach with several state-of-the-art methods including DRFI [33], Wco [26], IT [11], SF [50], GC [6], GMR [25], FT [2], MC [5], GB [1], HS [24], SVO [4], DSR [51], LEGS [34], BL15 [52] and BSCA [27].”
RGBD Salient Object Detection: A Benchmark and Algorithms (2014, methodology): “We select 12 state-of-the-art RGB saliency detection approaches, including the top 4 models ranked in the survey[6]: SVO[9], CA[14], CB[20], and RC[10], and 8 recently developed prominent methods: SF[34], LR[40], HS[44], MR[45], PCA[32], TD[37], GC[11], and DSR[28].”
SuperCNN: A Superpixelwise Convolutional Neural Network for Salient Object Detection (2015, methodology): “Like the previous work (Perazzi et al. 2012; Ming-Chng et al. 2013), we quantitatively evaluate the performances of these methods by measuring their precision-recall values, F-measure numbers, and mean absolute errors.”
Global Contrast Based Salient Region Detection (2015, unknown): “F”
Structure-Measure: A New Way to Evaluate Foreground Maps (2017, background): “As a speciﬁc example, here we focus on salient object detection models [4, 6, 7, 16], although the proposed measure is general and can be used for other purposes.”
Deep networks for saliency detection via local estimation and global search (2015, methodology): “We compare the proposed method (LEGS) with ten stateof-the-art models including SVO [4], PCA[26], DRFI [16], GC [6], HS [40], MR [41], UFO [17], wCtr [42], CPMCGBVS [23] and HDCT [18].”
Salient object detection via bootstrap learning (2015, methodology): “Left to right: (a) input (b) GS SP [34] (c) wCO [43] (d) LRMR [30] (e) GMR [38] (f) DSR [23] (g) XL13 [36] (h) HS [37] (i) RC-J [9] (j) GC [10] (k) SF [28] (l) Ours (m) wCO-bootstrapped (n) ground truth.”
BING: Binarized normed gradients for objectness estimation at 300fps (2014, unknown): “More recent research also tried to produce high quality saliency maps in a filtering based framework [46], using efficient data representation [12], or consider hierarchical structures [55].”
HCP: A Flexible CNN Framework for Multi-Label Image Classification (2014, background): “[31], [7], [46] are based on salient object detection, which try to detect the most attention-grabbing (salient) object in a given image.”
Saliency Optimization from Robust Background Detection (2014, unknown): “Among these, SF [17] and SIA [3] combine low level cues in straightforward ways; GS [26] and MR [29] use boundary prior; HS [28] and MR [29] use global optimization and are the best algorithms up to now.”
Visual saliency based on multiscale deep features (2015, background): “These saliency priors are either directly combined with other saliency cues as weights [8, 9, 20] or used as features in learning based algorithms [22, 23, 25].”