Pixel classification based color image segmentation using quaternion exponent moments.

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Image segmentation is a challenging problem due to its application-dependent nature and lack of a priori image structure information.
• Existing segmentation algorithms are often complex, leading to frequent undesired results.

Purpose of the Study:

• To propose a novel, efficient, and accurate color image segmentation method.
• To address the limitations of existing complex segmentation algorithms.

Main Methods:

• Pixel-level image features extracted using quaternion exponent moments (QEMs) to capture inter-channel color correlations.
• Twin Support Vector Machines (TSVM) classifier trained with Arimoto entropy thresholding for pixel classification.
• Segmentation of color images using the trained TSVM model.

Main Results:

• The proposed method effectively describes color image pixel content by considering correlations between color channels.
• The TSVM classifier demonstrates lower computation time and higher classification accuracy.
• Experimental results indicate promising segmentation performance compared to state-of-the-art methods.

Conclusions:

• The integration of QEMs and TSVM offers an effective approach for color image segmentation.
• The method provides a balance between computational efficiency and segmentation accuracy.
• This technique shows significant potential for various image segmentation applications.