Non-Rigid Multi-Modal Image Registration Using Cross-Cumulative Residual Entropy

Area of Science:

• Medical Imaging
• Computer Vision
• Image Processing

Background:

• Non-rigid registration aligns images with complex deformations.
• Existing methods struggle with multi-modality images and varying contrast.
• Information-theoretic measures are common but can be computationally intensive.

Purpose of the Study:

• To develop a robust and efficient non-rigid registration method for multi-modality images.
• To introduce cumulative residual entropy (CRE) as a novel information-theoretic measure for registration.
• To leverage B-spline representation for efficient deformation modeling.

Main Methods:

• Utilized cumulative residual entropy (CRE) as the core information-theoretic measure.
• Employed a tri-cubic B-spline representation for smooth, non-rigid transformations.
• Defined and maximized cross-CRE between images under transformation.
• Analytically computed the gradient of CRE for efficient parameter estimation.

Main Results:

• The proposed method demonstrates robustness across varying image contrast and brightness.
• Achieved faster convergence speeds compared to existing information-theoretic registration methods.
• Successfully registered images with large non-overlapping fields of view.
• Validated through experiments on both synthetic and real image datasets.

Conclusions:

• The combination of CRE and B-spline representation offers a robust and computationally efficient solution for non-rigid registration.
• The method is well-suited for challenging multi-modality registration scenarios.
• This approach advances the field of medical image analysis and computer vision.