Single-frame-based column fixed-pattern noise correction in an uncooled infrared imaging system based on weighted...

Area of Science:

• Infrared imaging technology
• Image processing and computer vision

Background:

• Uncooled infrared images commonly exhibit column fixed-pattern noise (FPN) due to readout circuit non-uniformity.
• Removing column FPN without causing stripe artifacts near strong edges is a significant challenge.

Purpose of the Study:

• To introduce a novel single-frame-based algorithm for accurate column FPN correction.
• To address the issue of stripe artifacts often introduced by existing destriping methods.

Main Methods:

• The algorithm employs two 1D filters: a weighted least-squares filter for edge-preserving horizontal smoothing and local weighted ridge regression for vertical refinement.
• Analysis of stripe artifact origins informs the algorithm's design for effective non-uniformity correction (NUC).

Main Results:

• The proposed algorithm successfully corrects column FPN while preserving image details.
• Demonstrated effectiveness in eliminating stripe artifacts, outperforming four state-of-the-art single-frame destriping methods.
• Validation through both simulated and real infrared image experiments.

Conclusions:

• The novel single-frame algorithm provides an effective solution for column FPN in uncooled infrared images.
• The method's ability to prevent stripe artifacts makes it a valuable advancement in infrared image processing.