Pixel clustering by adaptive pixel moving and chaotic synchronization

Area of Science:

• Computer Vision
• Image Processing
• Chaos Theory

Background:

• Traditional pixel clustering methods often struggle with ambiguous pixels and require pre-defined cluster numbers.
• Existing algorithms may lack robustness in segmenting complex image data.

Purpose of the Study:

• To propose a novel network of coupled chaotic maps for effective pixel clustering.
• To enhance image segmentation robustness and eliminate the need for pre-specifying the number of clusters.

Main Methods:

• A network of coupled chaotic maps is employed, where trajectories within a cluster synchronize.
• Desynchronization between trajectories of different clusters is utilized for segmentation.
• A pixel motion mechanism is introduced to improve cluster compactness and handle ambiguous pixels.

Main Results:

• The proposed method demonstrates synchronized chaotic map evolutions for intra-cluster pixels and desynchronized evolutions for inter-cluster pixels.
• The pixel motion mechanism enhances the model's robustness in classifying ambiguous pixels.
• The algorithm successfully performs pixel clustering without prior knowledge of the number of clusters.

Conclusions:

• The coupled chaotic map network offers a robust and flexible approach to pixel clustering.
• The method's ability to handle an unknown number of clusters and ambiguous pixels represents a significant advancement in image segmentation.