Matching by tone mapping: photometric invariant template matching.

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Nonlinear tone mappings present challenges for traditional pattern matching algorithms.
• Existing methods like normalized cross correlation (NCC) are limited to linear transformations.
• Mutual Information offers robustness but is computationally intensive.

Purpose of the Study:

• Introduce a novel, fast pattern matching scheme for nonlinear tone mappings.
• Generalize normalized cross correlation (NCC) to handle nonlinear mappings.
• Evaluate the discriminative power and robustness of the proposed method.

Main Methods:

• Developed Matching by Tone Mapping (MTM), a computational scheme for nonlinear tone mapping.
• Approximated tone mapping using piecewise constant/linear functions for efficiency.
• Compared MTM performance against normalized cross correlation (NCC) and mutual information.

Main Results:

• MTM achieves computational time comparable to fast NCC implementations.
• The MTM measure generalizes NCC and reduces to it for linear mappings.
• MTM demonstrates high discriminative capability and robustness to noise.

Conclusions:

• MTM provides an efficient and effective solution for pattern matching under nonlinear tone mappings.
• The method offers a practical alternative to computationally expensive techniques like mutual information.
• MTM represents a significant advancement in robust and fast pattern recognition for image processing.