Visual assessment of variable-resolution imagery

Area of Science:

• Image processing
• Computational vision
• Human visual system

Background:

• Human visual perception exhibits spatial inhomogeneities, with detail decreasing away from the center.
• Abnormalities in spatial discrimination can also affect visual processing.
• Existing models use eccentricity scaling functions to describe peripheral vision.

Purpose of the Study:

• To develop a technique for generating variable-resolution images.
• To match these images to the spatial properties of the human visual system.
• To explore efficient image sampling methods.

Main Methods:

• Generating images with linear distortion functions varying in low-pass characteristics.
• Using a forced-choice procedure to assess image indistinguishability from originals.
• Comparing distortion functions with established eccentricity scaling functions.
• Discussing locally band-limited spaces and developing an efficient sampling technique.

Main Results:

• Identified threshold distortion functions that characterize visual perception.
• Demonstrated that generated images can be visually equivalent to higher-information images.
• Proposed an efficient sampling technique based on locally band-limited spaces.

Conclusions:

• Variable-resolution image generation can simulate human visual processing.
• The described sampling technique offers efficient image creation with perceptual equivalence.
• This approach has potential applications in understanding visual perception and image compression.