Motion Blur Direction Estimation Using Radon Transform

The algorithm proceeds through the following steps: (1) First, motion blur is applied to the original image using a specified blur kernel (typically defined by length and angle parameters). The blurred image then undergoes Fourier transformation using functions like fft2() in MATLAB or np.fft.fft2() in Python, producing a frequency spectrum that reveals distinct parallel striations characteristic of motion blur. (2) The frequency spectrum of the motion-blurred image undergoes Canny edge detection, which involves Gaussian smoothing, gradient calculation, non-maximum suppression, and hysteresis thresholding. This process highlights the prominent parallel lines in the spectrum, which correspond to the motion blur direction. (3) The Canny edge detection result is then processed using Radon transform (implemented via functions like radon() in MATLAB's Image Processing Toolbox). The Radon transform computes line integrals at various angles, with the peak in the transform matrix indicating the dominant direction of the parallel lines, thus revealing the motion blur angle. To effectively implement this algorithm, developers should understand key concepts including motion blur kernel generation, fast Fourier transform (FFT) applications, Canny edge detection parameters (threshold settings), and Radon transform interpretation. Proper implementation requires careful parameter tuning for accurate direction estimation across different image types and blur conditions.