Visibility of objects in computer simulations of noisy micrographs

Area of Science:

• Image perception and visual science
• Signal processing and noise analysis

Background:

• Understanding object visibility in noisy images is crucial for various applications, including medical imaging and surveillance.
• Previous research established signal-to-noise ratio (SNR) as a key factor, but detailed analysis across different object characteristics was limited.

Purpose of the Study:

• To determine the visibility thresholds of objects obscured by random pixel noise.
• To extend existing SNR threshold measurements to various object sizes and shapes.
• To identify the key factors influencing object detectability in noisy visual scenes.

Main Methods:

• Conducted trials with human volunteers to mark objects in test images containing random pixel noise.
• Utilized test images with simple-shaped objects against a smooth, featureless background.
• Measured visibility thresholds in terms of signal-to-noise ratio (SNR) for different object sizes and shapes.

Main Results:

• Object visibility thresholds were successfully determined and correlated with SNR.
• For objects smaller than a 2-degree visual angle, visibility was found to depend on average intensity difference, noise level, and pixel count.
• Object shape did not appear to significantly influence visibility thresholds under the tested conditions.

Conclusions:

• The study confirms and extends the role of SNR in object visibility within noisy images.
• Visibility is primarily governed by image properties like intensity difference, noise, and object size, rather than its specific shape.
• These findings provide a quantitative basis for predicting object detectability in various imaging systems.