A coherent computational approach to model bottom-up visual attention

Abstract

Abstract-Visual attention is a mechanism which filters out redundant visual information and detects the most relevant parts of our visual field. Automatic determination of the most visually relevant areas would be useful in many applications such as image and video coding, watermarking, video browsing, and quality assessment. Many research groups are currently investigating computational modeling of the visual attention system. The first published computational models have been based on some basic and well-understood Human Visual System (HVS) properties. These models feature a single perceptual layer that simulates only one aspect of the visual system. More recent models integrate complex features of the HVS and simulate hierarchical perceptual representation of the visual input. The bottom-up mechanism is the most occurring feature found in modern models. This mechanism refers to involuntary attention (i.e., salient spatial visual features that effortlessly or involuntary attract our attention). This paper presents a coherent computational approach to the modeling of the bottom-up visual attention. This model is mainly based on the current understanding of the HVS behavior. Contrast sensitivity functions, perceptual decomposition, visual masking, and center-surround interactions are some of the features implemented in this model. The performances of this algorithm are assessed by using natural images and experimental measurements from an eyetracking system. Two adequate well-known metrics (correlation coefficient and Kullbacl-Leibler divergence) are used to validate this model. A further metric is also defined. The results from this model are finally compared to those from a reference bottom-up model.

Saliency Aggregation: Does Unity Make Strength?

“…To investigate this point, we select 8 state-of-the-art models (GBVS [3], Judd [14], RARE2012 [15], AWS [5], Le Meur [4], Bruce [7], Hou [8] and Itti [6]) and aggregate their saliency maps into a unique one. The following subsections present the tested aggregation methods.…”

“…Different algorithms are used to train the best way to combine together saliency maps. (c) Itti [6] (d) Le Meur [4] (e) GBVS [3] (f) Hou [8] (g) Bruce [7] (h) Judd [14] (i) AWS [5] (j) RARE2012 [15] …”

“…3.1 Performance of state-of-the-art models Figure 2 illustrates the performance of the 8 selected models (GBVS [3], Judd [14], RARE2012 [15], AWS [5], Le Meur [4], Bruce [7], Hou [8] and Itti [6]) over Bruce and Judd datasets. According to our results, we find that the top 2 models are GBVS and RARE2012, the top 4 models are GBVS, RARE2012, Judd and AWS.…”

“…The discrepancies are related to the quality of the prediction but also to the saliency map representation. Indeed some models output very focused saliency maps [3][4][5] whereas the distribution of saliency values is much more uniform in other models [6,7]. Others tend to emphasize more on the image edges [8], the color or luminance contrast.…”

Irregularity-based image regions saliency identification and evaluation

“…Correlation-based measures also were widely used, in which, the correlation between the saliency map generated by the saliency extraction algorithms and those that generated by using fixation data (Zhao & Koch, 2011), (Gide & Karam, 2012), (Mancas, 2008), (Parkhurst , et al, 2002), (Masciocchi, et al, 2009), (Rosin, 2009) and(Le Meur, et al, May 2006). 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…”