Evaluation of Geometric Depth Estimation Model for Virtual Environment
Puneet Kumar Sharma, Jan Harald Nilsen, Torbjørn Skramstad, Faouzi Alaya Cheikh · 2010
3-D virtual environment is a computer generated experience which gives us a feeling of presence in the environment. Objects displayed in virtual environment unlike the real world have no physical depth. Due to the distance between the eyes, the images formed on the retina are different, this facilitates our perception of depth. In the range of personal space, eyes converge at different angles to look at objects in different depth planes, known as convergence angle. Since we cannot get images of the scene viewed by the two eyes, the convergence angle cannot be calculated by standard photogrammetry principles such as triangulation. However, we can measure the point of focus(fixations) of the eyes on 2-D display plane, by using eye tracker. Each eye gets a different view of the virtual scene. Knowing the physical location of both eyes and their corresponding fixations, we can calculate the estimated depth using geometry. In this paper, first, we discuss the experiment setup and 3-D virtual scene used for depth estimation. Second, we evaluate the performance of the geometric model for depth estimation. Third, we discuss a histogram based filtering approach, for improving the performance of the geometric model. Results show that histogram based filtering improves the performance of the geometric model. 1