Biologically Inspired Hierarchical Contour Detection with Surround Modulation and Neural Connection

Area of Science:

• Computer Vision
• Computational Neuroscience
• Image Processing

Background:

• Contour detection is crucial for biological visual cognition.
• Conventional models struggle with contour detection in complex scenes.
• Biological vision mechanisms offer insights for improved contour extraction.

Purpose of the Study:

• To propose a hierarchical image contour extraction method based on biological vision.
• To overcome limitations of conventional contour detection models.
• To develop a computational model simulating early biological visual processing.

Main Methods:

• Simulating information processing in early biological visual system receptive fields.
• Developing a computational model integrating feedforward, lateral, and feedback neural connections.
• Hierarchical contour extraction leveraging perceptual characteristics like edges, shapes, and colors.

Main Results:

• The hierarchical contour detection model fits biological experimental characteristics.
• The model effectively detects salient contours in complex scenes.
• The model demonstrates superior suppression of unwanted textures compared to conventional methods.

Conclusions:

• The proposed model offers an effective approach to contour detection in complex visual scenes.
• The model's design, inspired by biological vision, enhances performance.
• This work advances computational models for image contour extraction.