An ASIC-chip for stereoscopic depth analysis in video-real-time based on visual cortical cell behavior

Area of Science:

• Computer Vision
• Computational Neuroscience
• Signal Processing

Background:

• Stereoscopic systems utilize slight differences between images from two viewpoints (disparities) to extract depth information.
• Existing methods for disparity computation often involve complex filtering and spatial analysis.

Purpose of the Study:

• To develop and present two novel algorithmic approaches for recovering disparity from stereo images.
• To formulate a disparity recovery method compatible with neural operations in the visual cortex.
• To introduce a real-time stereo-analysis approach inspired by auditory sound localization mechanisms.

Main Methods:

• Two algorithmic versions were developed, both comparing filter outputs from left and right image filtering.
• The first method employs Gabor filters and calculates spatial phase differences, aligning with neural processing.
• The second method transforms spatial disparity into the temporal domain using causal filters (resonators) for real-time analysis.

Main Results:

• Both algorithms successfully recover disparity by analyzing phase differences in filter responses.
• The temporal domain approach enables video real-time stereo-analysis, demonstrated by a developed FPGA-based PC-board.
• Hardware implementations (FPGA, ASIC) are discussed for practical, high-speed stereo processing.

Conclusions:

• Novel algorithms for disparity recovery offer insights into both computational neuroscience and practical computer vision applications.
• The temporal domain approach significantly advances real-time stereo-analysis capabilities for video sequences.
• The research bridges theoretical models of visual processing with efficient, hardware-accelerated implementations.