Saliency based illumination control for directing user attention in 3D scenes

Visual attention significantly affects how we perceive 3D environments and saliency is a part of visual attention expressing how likely a scene or item is to capture our attention due to its apparent features. Saliency relies upon geometry, shading, size, and other visual attributes of items. Illumination of a scene has an important impact on those visual properties; therefore, it influences the saliency distribution of a visual field. In this thesis, the saliency of objects will be recognized by different light locations. This work aims to control the saliency by manipulating the illumination parameters in a 3D scene. For this reason, given a sensible 3D scene, the light parameters that provide maximum saliency for the point of interest objects are investigated. In other words, we propose a method for task-aware automatic lighting setup. In this work, three approaches are proposed for 2D renderings of the 3D scene from specific viewpoints are considered and results are investigated according to saliency distribution under different light positions. For the first approach, a kitchen 3D scene that has objects understood by people is created with different light parameters and saliency results are investigated according to different light positions. Also, for this process, different saliency methods and calculations are used to get successful results and eye tracker testing is used to compare the results. For the second approach, a 3D scene with different light locations is created that has geometric objects not fully understood by people. Similarly, an eye-tracking experiment is conducted to compare relative saliency results and eye tracker-based saliency results. For the third approach, a 3D museum scene is created with light sources which maximize the visual attention of the target objects. In this approach, 360-degree renderings are taken to calculate the relative saliency, and an experiment which is based VR glass is conducted to compare the results. In each of our approaches, our aim is to investigate light positions where target objects are more conspicuous and confirm our results with salience values and eye tracking experiments.