Probability density estimation using isocontours and isosurfaces: applications to information-theoretic image...

Area of Science:

• Computer Vision
• Image Processing
• Geometric Methods

Background:

• Traditional image registration methods often struggle with noise and limited intensity resolution.
• Existing probability density estimation techniques can be sensitive to sampling and kernel selection.

Purpose of the Study:

• To develop a new geometric approach for probability density estimation in images.
• To improve affine image registration accuracy, particularly under challenging conditions.
• To extend the method for joint density estimation in image pairs and multiple images.

Main Methods:

• Representing images as piecewise-continuous surfaces rather than discrete pixels.
• Calculating probability density as the area between isocontours of the image surface.
• Applying the method to affine registration using mutual information.

Main Results:

• The proposed geometric method outperforms standard techniques like histograms and Parzen windows for affine registration.
• Demonstrated superior performance under fine intensity quantization and significant image noise.
• Successfully applied to simultaneous registration of multiple images and volume datasets.

Conclusions:

• The geometric approach offers a robust and effective alternative for probability density estimation in image analysis.
• This method simplifies density estimation by requiring only an image interpolant, avoiding arbitrary kernel functions or sampling.
• The technique shows significant promise for improving the accuracy and reliability of image registration tasks.