Restoration techniques for millimeter-wave images

A non-linear image restoration method has been developed by combining the advantages of the existing Lorentzian and Wiener filter techniques. It sharpens edges without introducing Gibbs ringing and restores the background without flattening it. An image is separated into features and background regions, the features are restored using the Lorentzian method and the background is sharpened using a Wiener filter. The Wiener filter is applied to the second derivative of the background to avoid ringing introduced by discontinuities where features have been removed. Also three pre-processing techniques are described that suppress fixed pattern noise, temporal noise and scan-lines from video data. The fixed pattern noise is suppressed by subtracting one frame of a moving image from another. Then the difference image is deconvoled with a function based on the translation of the image between each frame. Temporal noise is suppressed by calculating the displacement between frames and averaging the frames in their displaced position. Scan-lines are suppress by blurring the image in a direction perpendicular to the scan-lines and fitting the original image to the blurred image by adjusting gains and offsets. Examples of each method are provided.