Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Edge Detection based on Channel Attention and Inter-region Independence Test

Ru-yu Yan, Da-Qing Zhang

TL;DR

CAM-EDIT tackles noise-prone edge detection by marrying a parameter-efficient Channel Attention Mechanism with a statistically driven Independence Test. The CAM module enhances discriminative edge features, while EDIT suppresses uncorrelated noise by testing regional coordinate independence via $\,\chi^2$ and Fisher’s exact tests. On BSDS500 and NYUDv2, CAM-EDIT achieves state-of-the-art F-measures and demonstrates robustness to Gaussian noise, reflecting strong practical potential for industrial-grade edge extraction. The work introduces a compact, multi-scale edge feature fusion framework with statistically grounded refinement that improves edge fidelity and reduces artifacts in challenging scenes.

Abstract

Existing edge detection methods often suffer from noise amplification and excessive retention of non-salient details, limiting their applicability in high-precision industrial scenarios. To address these challenges, we propose CAM-EDIT, a novel framework that integrates Channel Attention Mechanism (CAM) and Edge Detection via Independence Testing (EDIT). The CAM module adaptively enhances discriminative edge features through multi-channel fusion, while the EDIT module employs region-wise statistical independence analysis (using Fisher's exact test and chi-square test) to suppress uncorrelated noise.Extensive experiments on BSDS500 and NYUDv2 datasets demonstrate state-of-the-art performance. Among the nine comparison algorithms, the F-measure scores of CAM-EDIT are 0.635 and 0.460, representing improvements of 19.2\% to 26.5\% over traditional methods (Canny, CannySR), and better than the latest learning based methods (TIP2020, MSCNGP). Noise robustness evaluations further reveal a 2.2\% PSNR improvement under Gaussian noise compared to baseline methods. Qualitative results exhibit cleaner edge maps with reduced artifacts, demonstrating its potential for high-precision industrial applications.

Edge Detection based on Channel Attention and Inter-region Independence Test

TL;DR

CAM-EDIT tackles noise-prone edge detection by marrying a parameter-efficient Channel Attention Mechanism with a statistically driven Independence Test. The CAM module enhances discriminative edge features, while EDIT suppresses uncorrelated noise by testing regional coordinate independence via and Fisher’s exact tests. On BSDS500 and NYUDv2, CAM-EDIT achieves state-of-the-art F-measures and demonstrates robustness to Gaussian noise, reflecting strong practical potential for industrial-grade edge extraction. The work introduces a compact, multi-scale edge feature fusion framework with statistically grounded refinement that improves edge fidelity and reduces artifacts in challenging scenes.

Abstract

Existing edge detection methods often suffer from noise amplification and excessive retention of non-salient details, limiting their applicability in high-precision industrial scenarios. To address these challenges, we propose CAM-EDIT, a novel framework that integrates Channel Attention Mechanism (CAM) and Edge Detection via Independence Testing (EDIT). The CAM module adaptively enhances discriminative edge features through multi-channel fusion, while the EDIT module employs region-wise statistical independence analysis (using Fisher's exact test and chi-square test) to suppress uncorrelated noise.Extensive experiments on BSDS500 and NYUDv2 datasets demonstrate state-of-the-art performance. Among the nine comparison algorithms, the F-measure scores of CAM-EDIT are 0.635 and 0.460, representing improvements of 19.2\% to 26.5\% over traditional methods (Canny, CannySR), and better than the latest learning based methods (TIP2020, MSCNGP). Noise robustness evaluations further reveal a 2.2\% PSNR improvement under Gaussian noise compared to baseline methods. Qualitative results exhibit cleaner edge maps with reduced artifacts, demonstrating its potential for high-precision industrial applications.
Paper Structure (21 sections, 13 equations, 10 figures, 8 tables, 1 algorithm)

This paper contains 21 sections, 13 equations, 10 figures, 8 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Traditional edge detection
  4. Methods of deep learning
  5. Method
  6. Channel Attention-Based Feature Extraction
  7. Dual-Convolution Layer
  8. ReLU Activation
  9. Pooling Layer
  10. Preliminary Edge Detection Algorithm
  11. Calculating Gradients and Membership Degrees
  12. Median Filter Coarsening and Morphological Smoothing of Images
  13. Edge Detection Method Based on Independence Test
  14. Experiment
  15. Dataset and edge evaluation metrics
  16. ...and 6 more sections

Figures (10)

  • Figure 1: Overview of the proposed CAM scheme.
  • Figure 2: Schematic diagram of convolution operation.
  • Figure 3: Overview of the proposed CAM-EDIT edge detection framework.
  • Figure 4: Extract continuous edge curves and randomly generated discrete points that can be regarded as noise points from the image.
  • Figure 5: The edge detection results of kernels of different sizes in the median filter. (a) $kernel=1$; (b) $kernel=3$; (c) $kernel=5$; (d) $kernel=7$.
  • ...and 5 more figures