Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Analysis of the Robustness of an Edge Detector Based on Cellular Automata Optimized by Particle Swarm

Vinícius Ferraria, Eurico Ruivo

TL;DR

This work analyzes the robustness of a cellular automaton–based edge detector optimized by Particle Swarm Optimization (PSO-CA) and augmented with transfer learning. By comparing radii r=1 and r=2 on the BSDS500 natural image set, it shows that expanding the search space does not improve edge detection under the tested threshold p, and transfer learning offers negligible gains. Across 10-fold cross-validation and category-wise experiments, PSO-CA generally outperforms the Canny detector in SSIM and PSNR, with the general model often matching or exceeding specialized models, though the landscape category remains challenging. The study highlights the adaptability of PSO-CA to image characteristics and suggests future work on optimization dynamics and dataset balance to further enhance robustness.

Abstract

The edge detection task is essential in image processing aiming to extract relevant information from an image. One recurring problem in this task is the weaknesses found in some detectors, such as the difficulty in detecting loose edges and the lack of context to extract relevant information from specific problems. To address these weaknesses and adapt the detector to the properties of an image, an adaptable detector described by two-dimensional cellular automaton and optimized by meta-heuristic combined with transfer learning techniques was developed. This study aims to analyze the impact of expanding the search space of the optimization phase and the robustness of the adaptability of the detector in identifying edges of a set of natural images and specialized subsets extracted from the same image set. The results obtained prove that expanding the search space of the optimization phase was not effective for the chosen image set. The study also analyzed the adaptability of the model through a series of experiments and validation techniques and found that, regardless of the validation, the model was able to adapt to the input and the transfer learning techniques applied to the model showed no significant improvements.

Analysis of the Robustness of an Edge Detector Based on Cellular Automata Optimized by Particle Swarm

TL;DR

This work analyzes the robustness of a cellular automaton–based edge detector optimized by Particle Swarm Optimization (PSO-CA) and augmented with transfer learning. By comparing radii r=1 and r=2 on the BSDS500 natural image set, it shows that expanding the search space does not improve edge detection under the tested threshold p, and transfer learning offers negligible gains. Across 10-fold cross-validation and category-wise experiments, PSO-CA generally outperforms the Canny detector in SSIM and PSNR, with the general model often matching or exceeding specialized models, though the landscape category remains challenging. The study highlights the adaptability of PSO-CA to image characteristics and suggests future work on optimization dynamics and dataset balance to further enhance robustness.

Abstract

The edge detection task is essential in image processing aiming to extract relevant information from an image. One recurring problem in this task is the weaknesses found in some detectors, such as the difficulty in detecting loose edges and the lack of context to extract relevant information from specific problems. To address these weaknesses and adapt the detector to the properties of an image, an adaptable detector described by two-dimensional cellular automaton and optimized by meta-heuristic combined with transfer learning techniques was developed. This study aims to analyze the impact of expanding the search space of the optimization phase and the robustness of the adaptability of the detector in identifying edges of a set of natural images and specialized subsets extracted from the same image set. The results obtained prove that expanding the search space of the optimization phase was not effective for the chosen image set. The study also analyzed the adaptability of the model through a series of experiments and validation techniques and found that, regardless of the validation, the model was able to adapt to the input and the transfer learning techniques applied to the model showed no significant improvements.

Paper Structure

This paper contains 18 sections, 10 equations, 12 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Edge Detection Task
  4. Two-Dimensional Cellular Automaton
  5. Cellular Automaton Edge Detector
  6. Particle Swarm Optimization
  7. Application of the Particle Swarm Optimization
  8. Cross-Validation in Categories
  9. Dataset
  10. Results
  11. Experimental Setup
  12. Experiments
  13. Metrics
  14. Discussion
  15. k-fold Cross-validation result analysis
  16. ...and 3 more sections

Figures (12)

  • Figure 1: von Neumann (left) and Moore (right) neighborhoods of radii 1 dumitru2021transfer (top) and 2 (bottom).
  • Figure 2: Representations of the proposed rule numbers of Moore's neighborhood $r = 1$dumitru2021transfer.
  • Figure 3: Definition of rule 362 for $r = 1$, rule composed by the combination of multiple cells dumitru2021transfer.
  • Figure 4: Representations of the proposed rule numbers of Moore's neighborhood $r = 2$.
  • Figure 5: Visualization of the pre-processed annotated edge maps of the BSDS500 dataset extracted with the probability threshold $p = 0.2$
  • ...and 7 more figures