On the usage of visual saliency models for computer generated objects

Abstract

Visual attention is a key feature to optimize visual experience of many multimedia applications. 2D visual attention computational modeling is an active research area considering the visualization of natural images on a conventional display. In this paper, we question the ability of such models to be applicable to single computer-generated objects rendered at different sizes (on a conventional display). We benchmark state of art visual attention models and investigate the influence of the viewpoint on those computational models applied on volumetric data and this to get a better understanding of how viewpoint information could be integrated into view-based approaches. To do so, a subjective experiment was conducted and a fine analysis was led using the variance analysis statistical method.

Perceptual Characterization of 3D Graphical Contents based on Attention Complexity Measures

“…Figure 1 shows two graphical objects belonging to two different semantic categories. More details about the dataset could be found here 1 .…”

“…The distance between the observer and the experimental display which is a computer monitor display with full HD resolution (1920 x 1080) was approximately 110 cm. This distance was defined in such a way as to guarantee an accurate recording while also ensuring comfortable viewing for the observer 1 . The total time of the experiment was 15 minutes including vision check, calibration and 21 x 4 = 84 stimuli visualization during 3 seconds each (which is sufficient to cover both bottom-up and topdown visual attention behaviors in a balanced manner).…”

“…From the collected gaze data, we derived fixations using three thresholds 1 : motion ( • ), velocity ( • / ), and acceleration ( • / 2 ) . Afterwards, we computed ground-truth (GT) saliency maps, representing the probability of gazing for each pixel recorded from human data.…”

“…Several metrics exist to measure the alignment between model predictions and human eye fixations [5]. Such study was addressed on computer generated contents in [1]. In this paper saliency models performance is not detailed.…”

Towards Visual Saliency Computation on 3D Graphical Contents for Interactive Visualization

“…Afterwards, we apply a 2D view-based saliency model on the 2D projections of the 3D object. In this work, we considered Salicon model [19] as it showed the highest performances when computed on computer generated contents [20]. The 2D saliency map represents a probability map, and the saliency value at each location indicates the chances of how likely people paying attention there.…”