Performance consequences of two types of stereo picture compression

Area of Science:

• Computer Vision
• Human Perception
• Image Processing

Background:

• Stereo vision relies on processing information from two images to perceive depth.
• Existing theories of depth perception include information fusion and suppression.
• Digital image compression techniques can reduce data size but may impact perceptual quality.

Purpose of the Study:

• To evaluate two distinct stereo picture compression algorithms.
• To assess the impact of compression strategies on depth and object decision-making tasks.
• To determine which compression approach aligns better with human depth perception theories.

Main Methods:

• Developed two stereo compression algorithms: one with equal information reduction, another with deep compression of one image.
• Recruited human subjects to perform depth and object identification tasks on compressed stereo images.
• Compared performance on compressed images against uncompressed control conditions.

Main Results:

• Subjects could perform depth and object decisions on compressed images, though with reduced accuracy compared to uncompressed images.
• Performance in both tasks was significantly better when information reduction was uneven between the two stereo images.
• The findings indicate a preference for compression strategies that heavily reduce information in one image.

Conclusions:

• The results support the suppression theory of depth perception over the fusion theory.
• Uneven information compression in stereo images appears more perceptually viable for human observers.
• This research has implications for developing more effective stereo image compression methods.