A Novel Scheme Based on the Diffusion to Edge Detection.

Area of Science:

• Computer Vision
• Image Processing
• Computational Physics

Background:

• Current edge detection methods rely on gradient-based techniques or resource-intensive machine learning models.
• Machine learning approaches, like deep neural networks, require extensive labeled datasets and significant computational resources for training.
• Traditional methods often struggle with variations in scale and lighting conditions.

Purpose of the Study:

• To present a novel edge detection scheme based on the physical law of diffusion.
• To offer an alternative to gradient-based and machine learning methods for edge detection.
• To develop a method that preserves image structural information effectively.

Main Methods:

• The proposed scheme models edge detection as energy diffusion within image media.
• It utilizes the characteristic function of diffusion to extract energy information.
• This approach avoids treating edges solely as grayscale or color gradients.

Main Results:

• The diffusion-based scheme successfully preserves the structural information of images.
• It demonstrates a unified mathematical framework applicable to various conditions, including multiscales and different lighting.
• The method exhibits low computational complexity compared to existing techniques.

Conclusions:

• The diffusion-based edge detection scheme provides a robust and efficient alternative.
• Its foundation in physical laws offers a unified approach for diverse image conditions.
• This method reduces the reliance on large datasets and extensive training resources common in machine learning.