A Visual-Attention Model Using Earth Mover's Distance-Based Saliency Measurement and Nonlinear Feature Combination

Lin, Yuewei; Tang, Yuan Yan; Fang, Bin; Shang, Zhaowei; Huang, Yonghui; Wang, Song

doi:10.1109/tpami.2012.119

Cited by 30 publications

(20 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this model, elementary features, e.g., color and luminance computed from different scales, are integrated using a center-surround operator to generate the saliency map, in which visually salient points are highlighted, as the prediction of fixations. After that, a number of fixation prediction models are proposed (e.g., [12,25]). A comprehensive survey on the fixation prediction models can be found in [5].…”

Section: Related Workmentioning

confidence: 99%

HARF: Hierarchy-Associated Rich Features for Salient Object Detection

Zou

Komodakis

2015

2015 IEEE International Conference on Computer Vision (ICCV)

View full text Add to dashboard Cite

The state-of-the-art salient object detection models are able to perform well for relatively simple scenes, yet for more complex ones, they still have difficulties in highlighting salient objects completely from background, largely due to the lack of sufficiently robust features for saliency prediction. To address such an issue, this paper proposes a novel hierarchy-associated feature construction framework for salient object detection, which is based on integrating elementary features from multi-level regions in a hierarchy. Furthermore, multi-layered deep learning features are introduced and incorporated as elementary features into this framework through a compact integration scheme. This leads to a rich feature representation, which is able to represent the context of the whole object/background and is much more discriminative as well as robust for salient object detection. Extensive experiments on the most widely used and challenging benchmark datasets demonstrate that the proposed approach substantially outperforms the state-of-theart on salient object detection.

show abstract

Section: Related Workmentioning

confidence: 99%

HARF: Hierarchy-Associated Rich Features for Salient Object Detection

Zou

Komodakis

2015

2015 IEEE International Conference on Computer Vision (ICCV)

View full text Add to dashboard Cite

show abstract

“…Area Under the Curve was used for this purpose in which the saliency map is converted to binary image and then the AUC is calculated and compared to the AUC extracted from the ground truth data (Lin, et al, 2013) (Gide & Karam, 2012), (Zhao & Koch, 2011), (Erdem & Erdem, 2013), and (Kim & Milanfar, 2013). Tatler supported the idea of empirical evaluation of the saliency since they were interested more in mimicking the natural behaviour of the humans' eyes (Tatler, et al, 2011).…”

Section: Saliency Evaluationmentioning

confidence: 99%

“…Some other techniques were proposed, such as, Least Square Index (Zhao & Koch, 2011), Earth Mover's Distance (Zhao & Koch, 2011), (Lin, et al, 2013), (Judd, et al, January 13, 2012), (Pele & Werman, 2008), (Lin, et al, 2010) and (Rubner, et al, 2000), Receiver…”

Section: Saliency Evaluationmentioning

confidence: 99%

Irregularity-based image regions saliency identification and evaluation

Al-Azawi

Yang

Istance

2014

Multimed Tools Appl

View full text Add to dashboard Cite

IntroductionSalient points, objects, or regions extraction is still a challenging topic of research and its definition is still very general. Different literature have defined the saliency in different ways. In general and based on human vision system HVS, the salient object is the object that may capture the attention or attract the HVS. HVS uses two stages to identify the objects, preattentive and attentive stages (Kadir & Brady, 2001).In pre-attentive stage, the local regions of image that present spatial discontinuity are detected. In the attentive stage, relationships between these regions are found, and grouped together. The rest of the paper will be organized as follows: previous work is presented in section 2. In Section 3, we will explain the proposed saliency extraction technique based on the irregularity of the regions in an image. In section 44, a new automatic thresholding technique will be proposed. Points clustering and region merging will be discussed in section 5. Section 6 will review the saliency evaluation technique and introduce a new method that is suitable to the application at hand. In section 7 we will discuss the experimental results and the optimum window size selection. Finally, conclusions will be derived in section 8. 2 Previous WorkWavelet was used by many authors such as in (Loupias, et al., 2000), (Tian, et al., 2001), (Song, et al., 2006), (Lin & Yang, 2007), and (Arivazhagan & Shebiah, 2009) In this paper, we shall introduce a robust and general saliency extraction method that is fully automated and does not need any intervention from the user. In addition, it does need any pre-knowledge about the image. Irregularity as a Measure of SaliencyIn this Section, we shall discuss the proposed technique. First, we shall define the saliency as any irregular region in the image with nature differs from the nature of the image as a whole; for example, a region with texture in a uniform region, or vice versa, thus we shall define irregularity as the main measure of saliency.Let ℐ( , ) be the image intensity at position ( , ), where = 1, 2, … . , , = 1,2, … , , with and are the image width and height respectively. The image is represented as a set of pixels and is defined as:Where p xy is the pixel's value at location ( , ) and ℕ is the set of natural numbers. Furthermore, we shall define the set ℙ = { | ⊆ } as the power set of .According to the application at hands, the main constrain is that; all the elements of a given subset should belong to a connected region which means they should be in a given neighbouring area.The sub-image, which we shall refer to as a region, is a subset of the image with elements that belong to the same region. The size of this sub-image is × ℎ.Considering that the regions are disjoint, they can be defined as:Assuming that there is no overlapping between the adjacent regions, the total number of regions is equal to × ℎ . In most cases there should be overlapping between the moving windows, if we denote the overlapping value by , then the total numbers of wind...

show abstract

“…Multi-scale analysis is conventional and useful for visual saliency detection, and it is widely used in many literatures [9,13,14]. For a same object, attention on small scale image focuses on a whole object, while attention on large scale image cares more about the local details.…”

Section: Bicubic Interpolationmentioning

confidence: 99%

Visual saliency detection based on multi-scale and multi-channel mean

Sun

Tang

Zhang

2014

Multimed Tools Appl

View full text Add to dashboard Cite

This paper proposes an effective method for visual saliency detection based on multi-scale and multi-channel mean. 2-D wavelet transform is used to decompose and reconstruct image. Bicubic interpolation algorithm is applied to narrow the filtered image in multi-scale. We take the distances between the narrowed images and the means of their channels as saliency values, and we only reserve part values which are not less than the mean saliency of the given image. Bicubic interpolation algorithm is applied again to amplify the images in multi-scale, and then the saliency map is calculated by adding the amplified images. Finally, linear normalization is employed to obtain the final saliency map. Experimental results show that the proposed method outperforms 9 state-of-the-art methods both on the definition and accuracy of salient detection.

show abstract

A Visual-Attention Model Using Earth Mover's Distance-Based Saliency Measurement and Nonlinear Feature Combination

Cited by 30 publications

References 48 publications

HARF: Hierarchy-Associated Rich Features for Salient Object Detection

HARF: Hierarchy-Associated Rich Features for Salient Object Detection

Irregularity-based image regions saliency identification and evaluation

Visual saliency detection based on multi-scale and multi-channel mean

Contact Info

Product

Resources

About