Detection of Targets in Filtered Noise: Whitening in Space and Spatial Frequency

Table of Contents

1. background: functions to create specific types of noise backgrounds.
2. experiment: experiment data of detecting 1.5-cpd and 3-cpd targets in 1/f noise.
3. figure_data: results of analysis used to plot figures in our corresponding paper (in submission).
4. filter_operation: filter operation used in the template matching models.
5. matching: template matching models juggling whitening in space, whitening in spatial frequency, eye filtering and positional uncertainty.
6. mathematics: mathematical functions used in the project. Special thanks to Abhranil Das for the Matlab package to integrate and classify normal distributions.
7. simulation: simulation results of two experiments mentioned above and the exploration of model performance in natural images.

Function / Purpose

This repository is a summary of my research project under the supervision of Wilson Geisler. It provides optimal and sub-optimal model observers for detecting deterministic targets in wide sense stationary 2D noise. These models are also very efficient in non-stationary noises, such as the natural images.

Reference

For more details of this project, welcome to read the corresponding peer-reviewed article:

Zhang, A., & Geisler, W. S. (2022). Detection of targets in filtered noise: whitening in space and spatial frequency. JOSA A, 39(4), 690-701.

Correction: The luminance of the screen outside the background patch was set to the mean luminance of the background patches, which was always 46 $\textrm{cd/m}^2$.

Contact

Email: anqizhang@utexas.edu

LinkedIn: www.linkedin.com/in/anqi-work