Modelling biological depth perception in binocular vision: the local disparity estimation

Area of Science:

• Computer Vision
• Computational Neuroscience
• Artificial Intelligence

Background:

• The correspondence problem is crucial for stereopsis in binocular vision.
• Existing models, like Marr and Poggio's, provide a basis for comparison.
• Biologically inspired approaches offer novel solutions for visual processing.

Purpose of the Study:

• To present a novel approach for solving the correspondence problem in binocular vision.
• To compute local horizontal disparity maps using a biologically inspired algorithm.
• To develop a computational framework for stereopsis models and integrate disparity maps with other visual cues.

Main Methods:

• Implementation of two stereopsis models: one biologically inspired (modeling striate cortex cells) and the classical Marr and Poggio model.
• Development of a computer application for testing computational models of stereopsis.
• Testing on random-dot stereograms and real image pairs.

Main Results:

• The biologically inspired model demonstrated effectiveness in computing disparity maps.
• Simulation results provided a basis for comparing the novel approach with the classical model.
• The developed application served as a functional framework for stereopsis research.

Conclusions:

• The biologically inspired algorithm offers a viable approach to the correspondence problem in binocular vision.
• The developed computational framework facilitates the testing and integration of stereopsis models.
• Further research can leverage this framework for advancing computational models of vision.