Compound filter using the circular harmonic expansion

Area of Science:

• Optical signal processing
• Image processing
• Pattern recognition

Background:

• Traditional filtering methods struggle with rotation invariance and noise suppression.
• Accurate target identification requires robust correlation techniques.
• Existing methods often yield broad correlation peaks, hindering precise localization.

Purpose of the Study:

• To develop a compound rotation-invariant filtering technique for sharp target correlation peaks.
• To suppress random noise and unwanted target correlations effectively.
• To maximize the signal-to-noise ratio (SNR) of the filter output.

Main Methods:

• Utilizing a filter bank of circular harmonic filters.
• Weighting and summing output image moduli for the final image.
• Constraining weights to achieve a peak value of 1 for the desired target.
• Minimizing noise measures via weight optimization, solving a least-squares problem.

Main Results:

• Achieved sharp cross-correlation peaks for desired targets.
• Successfully suppressed random and deterministic noise.
• Maximized the signal-to-noise ratio (SNR) of the compound filter.
• Demonstrated unique weight solutions through a classical least-squares approach.

Conclusions:

• The developed compound filter technique offers robust rotation-invariant target detection.
• The method effectively balances target peak sharpness with noise reduction.
• Potential for optical, digital, and hybrid implementations with discussed performance limitations and extensions.