Point spread functions and deconvolution of ultrasonic images

Area of Science:

• Non-destructive testing
• Ultrasonic imaging
• Signal processing

Background:

• Ultrasonic pulse-echo C-scan imaging is crucial for non-destructive testing.
• Image quality is often limited by system blurring, necessitating restoration techniques.
• Linear System Theory (LST) provides a framework for understanding and modeling imaging systems.

Purpose of the Study:

• To investigate the deconvolution of ultrasonic C-scan images using a derived point spread function (PSF).
• To establish a robust model for the pulse-echo imaging process based on LST and wave equation formulation.
• To compare various deconvolution algorithms for optimal image reconstruction.

Main Methods:

• Derivation of an analytic formula for the PSF of planar transducers.
• Evaluation of numerical and analytic approximation schemes for PSF calculation.
• Comparison of deconvolution algorithms: Wiener filter, ForWaRD, and Richardson-Lucy with total variation regularization.
• Validation using simulated and measured C-scan images.

Main Results:

• The LST assumptions and derived PSF formula accurately model the pulse-echo imaging process.
• The Richardson-Lucy algorithm with total variation regularization yielded the best image reconstruction results.
• A simplified far-field approximation for the PSF is effective at distances beyond twice the near-field distance.

Conclusions:

• Deconvolution with a derived PSF is an effective method for restoring ultrasonic C-scan images.
• The Richardson-Lucy algorithm offers superior performance for ultrasonic image reconstruction.
• Simplified PSF approximations can be utilized in specific imaging scenarios, optimizing computational efficiency.