Image Thresholding Segmentation on Quantum State Space.

Area of Science:

• Quantum Computing
• Image Processing
• Computer Vision

Background:

• Traditional image segmentation relies on thresholding techniques, which can be computationally intensive and lack explicit physical interpretations.
• Existing methods like Otsu's method provide a basis for thresholding but can be improved in terms of efficiency and interpretability.
• Quantum mechanics offers a novel framework for representing and processing image data.

Purpose of the Study:

• To develop precise and efficient image segmentation methods using quantum state space.
• To introduce new image encoding techniques for optimal thresholding.
• To provide a more explicit physical interpretation of image thresholding through quantum mechanics.

Main Methods:

• Representing pixel intensities and their probability distributions using state vectors and density matrices in quantum mechanics.
• Proposing global quantum entropy maximization (GQEM) with an objective function equivalent to Otsu's method.
• Developing quantum lossy-encoding-based entropy maximization (QLEEM) to avoid optimal threshold searching by utilizing eigenvalues of density matrices.

Main Results:

• The QLEEM algorithm implicitly determines the thresholding level upper bound via eigenvalues.
• Proposed methods retain local image information while maximizing inter-class separability, enhancing structural characteristics.
• Experimental results demonstrate competitive thresholding quality and the fastest computation speed compared to state-of-the-art methods.

Conclusions:

• The novel quantum-based image thresholding methods (GQEM and QLEEM) offer significant improvements in efficiency and interpretability.
• QLEEM provides a faster approach by eliminating the need for optimal threshold searching.
• These quantum-inspired techniques effectively balance local information preservation and inter-class separability for superior image segmentation.