Deconvolution of multiple images of the same object.

Area of Science:

• Image Processing and Computer Vision
• Signal Processing
• Computational Imaging

Background:

• The image deconvolution problem has been a long-standing challenge in image processing.
• Previous methods, like Berenstein's, did not fully account for real-world factors such as sensor noise.
• A more realistic approach is needed for robust image restoration.

Purpose of the Study:

• To develop a novel image deconvolution method using data from multiple sensors with varying point spread functions.
• To incorporate image sensor noise and adaptive restoration into the deconvolution framework.
• To estimate signal and noise parameters directly from observed noisy signals.

Main Methods:

• Utilized a multi-sensor image acquisition strategy with different point spread functions.
• Developed an adaptive restoration algorithm that estimates signal and noise parameters from noisy data.
• Analyzed the impact of the number of sensor channels and signal-to-noise ratio on restoration accuracy.

Main Results:

• Demonstrated that arbitrary restoration accuracy is achievable through optimal selection of sensor channels and signal-to-noise ratios.
• Showcased the effectiveness of the adaptive restoration approach in handling sensor noise.
• Extended the method to scenarios where true images across different sensor channels are not identical.

Conclusions:

• The proposed multi-sensor deconvolution approach offers a significant advancement over existing methods.
• Adaptive restoration with noise consideration and parameter estimation from noisy signals is crucial for realistic image deconvolution.
• The method provides a flexible and powerful tool for achieving high-fidelity image restoration in complex scenarios.