Detecting motion blur based on ratio of image projection and camera shutter sequence transform

摘要

A real-time system and process comprising obtaining real-world object digital images from a shutter camera and estimating image motion blur using a ratio of values obtained from the projection of a 2D spectrum of the image and a Fourier transform of the shutter sequence used in obtaining the image. The projection may be obtained using a Radon transform. Estimating motion blur may use positions at peaks and valleys of the projection, including the global minimum. Also disclosed is a system and process comprising determining a Radon transform of the power spectrum of a blurred image, determining the blur direction by detecting the direction displaying the most variance, applying a smoothing filter to remove noise, computing key locations at the peaks and troughs of the spectrum, computing a modulation transfer function (MTF) of the shutter sequence and key locations of the MTF, and computing image blur extent from the weighted sum of the ratios of the key locations from the smoothed spectrum and the MTF shutter sequence.