Image preprocessing to enhance phase correlation of featureless images

Area of Science:

• Image processing
• Computer vision
• Remote sensing

Background:

• Phase correlation is effective for feature-rich image registration but struggles with featureless images.
• Noise in phase correlation for featureless imagery is not explained by current models.
• Applications like cloud tracking require robust registration of low-texture images.

Purpose of the Study:

• To develop an improved mathematical model for phase correlation in featureless images.
• To propose and evaluate novel image pre-processing techniques for enhancing phase correlation.
• To address the prominent noise component observed in featureless image registration.

Main Methods:

• An additional term was incorporated into the mathematical description of phase correlation.
• Expressions for phase correlation were derived for various scenarios, including featureless imagery.
• Several image pre-processing methods were developed and tested.

Main Results:

• The proposed mathematical modifications and pre-processing techniques led to significant noise reduction.
• Phase correlation performance was dramatically enhanced for featureless images, demonstrated on sky imagery.
• The new approach effectively overcomes limitations of traditional phase correlation in low-texture environments.

Conclusions:

• The enhanced mathematical framework and pre-processing strategies provide a robust solution for phase correlation in featureless imagery.
• This work offers practical improvements for applications such as satellite imagery analysis and meteorological tracking.
• The findings pave the way for more accurate image registration in challenging visual conditions.