HiSharpen uses a linear frequency domain filter to remove the minor blur caused by the optical system of the HiRISE camera. This blur is estimated from prelaunch measurements of the camera's point spread function (PSF) supplied by Alan Delamere of Ball Aerospace. The program accepts as input a HiRISE subimage of dimensions 2**n square. The program determines which filter (Red, BG, IR) was used for acquiring the HiRISE image and then selects the corresponding PSF and scales it to match the dimensions of the input image. First, the Fourier transforms of both the image and the PSF are calculated using the Fast Fourier Transform (FFT) algorithm. Next, the transform of the image is divided by the transform of the PSF (the Modulation Transfer Function, or MTF). The quotient is then transformed back to the spatial domain by an inverse FFT. Note that no pre-whitening or other noise suppression is done by HiSharpen. Random noise and pixel-to-pixel variations will be accentuated by HiSharpen unless steps have been taken to reduce the noise in the input image prior to sharpening. Note also that the FFT algorithm assumes that the input image is periodic, i.e. the left side of the image is repeated when the right edge of the image is reached. HiSharpen may therefore produce severe ringing in input images with low spatial frequency variations.

This is hisharpen being run on a cropped version of a hirise cube, the cropped area has dimensions of 256 by 256.