Optimized light source spectral power distribution for RGB camera based spectral reflectance recovery

Area of Science:

• Color Science
• Computational Imaging
• Optical Engineering

Background:

• Accurate spectral reflectance recovery is crucial for various applications, including digital archiving, material analysis, and computer vision.
• The accuracy of spectral estimation is influenced by the spectral estimation algorithm, camera characteristics, filters, and the illumination source spectrum.

Purpose of the Study:

• To present and compare different light source spectrum optimization methods for spectral estimation.
• To evaluate the performance of various spectral estimation algorithms in conjunction with optimized light sources for reflectance recovery.
• To assess the impact of optimized spectra on camera spectral sensitivities.

Main Methods:

• Optimization techniques including Monte Carlo (MC), particle swarm optimization (PSO), and multi-population genetic algorithm (MPGA) were employed to optimize light source spectra.
• Optimized spectra were compared against standard illuminants (D65, D50, A) and LED sources in both simulated and real-world experiments.
• Spectral estimation algorithms were applied to recovered reflectance data.

Main Results:

• The multi-population genetic algorithm (MPGA) exhibited superior performance in spectral estimation accuracy compared to other methods.
• Optimized light source spectra, when combined with advanced spectral estimation algorithms, led to more accurate spectral reflectance estimation.
• Camera spectral sensitivities, when weighted by optimized spectra, showed a tendency towards mutual orthogonality, potentially improving estimation robustness.

Conclusions:

• The choice of spectral estimation algorithm and the optimization of the illumination source spectrum are critical factors for accurate spectral reflectance recovery.
• MPGA offers a promising approach for optimizing light source spectra in spectral estimation tasks.
• Further research into the relationship between optimized spectra and camera sensor characteristics can enhance reflectance estimation accuracy.