Related Experiment Video
Updated: Jun 27, 2026
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
Automated design of image operators that detect interest points
Leonardo Trujillo1, Gustavo Olague
1Proyecto Evovisión, Departamento de Ciencias de la Computación, División de Física Aplicada, Centro de Investigación Científica y de Educación Superior de Ensenada, Km. 107 Carretera Tijuana-Ensenada, 22860, Ensenada, BC, México. trujillo@cicese.mx
Evolutionary computation, specifically genetic programming (GP), automatically designs effective image operators for interest point detection. This approach yields competitive and novel operators, advancing computer vision feature extraction.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Image Processing
Background:
- Interest point detection is crucial for computer vision tasks like object recognition and image indexing.
- Traditional methods often rely on manually designed image operators.
- Evolutionary computation, particularly genetic programming (GP), offers an alternative for automated operator design.
Purpose of the Study:
- To explore the use of evolutionary computation (GP) for synthesizing low-level image operators for interest point detection.
- To automatically generate operators that are competitive with state-of-the-art methods.
- To provide new perspectives on feature extraction through unconventional operator designs.
Main Methods:
- The study frames the design of image operators as an optimization/search problem.
- Genetic programming (GP) is employed to automatically synthesize these operators.
- Fitness evaluation considers geometric stability and global separability of detected points.
Main Results:
- The proposed GP approach successfully synthesized novel image operators.
- Synthesized operators demonstrated performance competitive with existing state-of-the-art designs.
- Fifteen new operators were generated, including both improved conventional and unconventional designs.
Conclusions:
- Evolutionary computation provides a viable and effective method for synthesizing specialized image operators.
- GP-based synthesis can lead to operators that match or exceed human-designed counterparts.
- The generation of unconventional operators opens new avenues for research in feature extraction.