武汉理工大学数字图像处理英文课件 Chapter7（5）.pptVIP

2D Discrete Fourier Transform Prepare Image for FFT Steps in Generating Fourier Transform Image Simple Example Comparison to Theoretical Result Normalization Image Comparison Matlab Function fftprep Matlab Function reduce * Fourier (1768-1830) Fourier’s Transform 1. Start with a simple piece of artwork (55 x 55). smile = imread(smile.tif); 2. Embed it in a 64 x 64 array. smile1 = fftprep(smile); 3. Reduce the size to keep our images small (optional). smile2 = reduce(smile1); 4. Embed this image in a 128 x 128 array (optional). smile3 = fftprep(smile2); 1. Remap the origin to the upper left corner smile3 = fftshift(smile3); 2. Generate the Fourier transform smile4 = fft2(smile3); 3. Construct the square modulus = (real*real+imag*imag), scale the results to a maximum value of unity, and remap the origin back to the center of the image. sqmod = smile4.*conj(smile4); Sqmod = fftshift(sqmod)/max(sqmod(:)); 4. Use logarithmic enhancement to see details. The default stretches 4 decades of values. logim(sqmod); complex-valued image Before applying the Fourier Transform to general images, we really should apply it to a case for which we know the answer. ???????????????? ????????????? ???????????????? We are actually displaying the square modulus of the Fourier transform or (somb(x))2. The input image is a circular disk with a radius of 4 pixels centered in a 128 x 128 array. The output image is the square modulus of the resulting Fourier transform. A cross-section of the output image is shown below. The marked data points are taken from a horizontal cross-section of the output image. The solid curve is the ideal function, which was generated from somb(r/16).^2. The factor of 16 converts pixel distance to the appropriate spatial-frequency distance. If the input disk is 1 unit in diameter and its radius is 4 pixels, then 8 pixels = 1 unit. The pixel spacing is then The period X of the discrete transform is The fundamental frequency xo is the inverse of the