Quantifying the benefits of image restoration

To quantify the benefits of applying image restoration algorithms to passive millimetre imagery a synthetic scene was generated and restored and the image quality assessed. The synthetic scene consisted of a three bar fan pattern and a large rectangular black area. This pattern was blurred with a Gaussian point spread function, and two- percent noise added to the simulated imagery. The blurred noisy image was then passed through a series of image restoration algorithms. A comparison was made between the MAP and the Wiener image restoration algorithms in terms of three quantities calculated before and after restoration, the rms noise, the point spread function and Gibbs ringing. The images were then compared to images taken using the DERA MITRE imager of an actual three bar fan pattern constructed from Stirling board. It was found that the performance of the algorithms on real imagery was limited by fixed pattern noise generated by the imaging process. The real data was improved further by applying a two-point correction based on features in the image and although an improvement was visible, the restoration was not as good as that observed for the synthetic data.