Related Experiment Video
Updated: May 29, 2026
07:45
Assessing Binocular Central Visual Field and Binocular Eye Movements in a Dichoptic Viewing Condition
Published on: July 21, 2020
Multiple-window parallel adaptive boundary finding in computer vision
1MEMBER, IEEE, Division of Engineering, Brown University, Providence, RI 02912.
IEEE Transactions on Pattern Analysis and Machine Intelligence
|August 27, 2011
Summary
This study introduces parallel multiple-window algorithms for accurate object boundary estimation in noisy images. The method uses dynamic programming within a maximum likelihood framework for robust boundary detection and reconstruction.
Area of Science:
- Image processing and computer vision
- Computational imaging
- Pattern recognition
Background:
- Accurate object boundary estimation is crucial in various imaging applications, including medical scans and infrared imaging.
- Noisy image data and complex object shapes present significant challenges to traditional boundary detection methods.
- Existing algorithms often struggle with high variability and internal object structures.
Purpose of the Study:
- To develop and evaluate a novel parallel multiple-window algorithm for robust object boundary estimation in noisy images.
- To integrate boundary finding within a unified maximum likelihood estimation framework.
- To improve the accuracy and efficiency of boundary detection in diverse imaging scenarios.
Main Methods:
- Partitioning the image field into an array of rectangular windows for parallel processing.
- Employing dynamic programming-based boundary finders within each window.
- Seaming boundary segments from individual windows to reconstruct global object boundaries.
- Utilizing a maximum likelihood estimation framework for overall boundary estimation.
Main Results:
- Demonstrated effectiveness of the parallel multiple-window approach for estimating highly variable object boundaries.
- Successful integration of dynamic programming with maximum likelihood estimation for precise boundary localization.
- Validation of the F-test for efficient boundary presence detection within windows.
- Analysis showing improved boundary recognition probability using coarse pixels with chi-square or F-tests.
Conclusions:
- The proposed parallel multiple-window boundary estimation algorithm offers a robust solution for noisy and complex imagery.
- The maximum likelihood framework provides a principled approach to integrating local boundary findings into global structures.
- Computational efficiency is enhanced through techniques like the F-test and the use of coarse pixels, improving practical applicability.
