为什么在 matlab 中使用逆傅立叶变换 ifft2() 时图像变暗？

Question

I believe there's something wrong when I convert it back using ifft2(). I am trying to use the ideal low-pass filter by creating a matrix L that has a circle with 1's with the radius 50. It would be great if you guys can tell me what the issue is. Thanks!

I0 = imread('image.png');
g = I0(:,:,1);
Rg = imref2d(size(g));


G = fft2(g);
AG1 = log(1+abs(G));
MaxVal = max(max(AG1));

AG2 = uint8(255*(AG1/MaxVal));
SAG2 = fftshift(AG2);


%zero matrix L with a circle with value 1's with radius 50 
[xGrid,yGrid] = meshgrid(1:400,1:400);
L = sqrt((xGrid - 200).^2 + (yGrid - 200).^2) <= 50;

U = uint8(double(SAG2).*L);
a=ifft2(U);


u = uint8(real(ifft2(U)));

imshow(u,'InitialMagnification',300);

Answer 1

You are taking a whole bunch of unnecessary actions. As mentioned in the comments, all the int8 castings are redundant. You are also missing an ifftshift in the calculation of u . Here is a short version that does the low-passing you want:

I0 = imread('your_file.png');
g = I0(:,:,1);

G = fft2(g);

SAG2 = fftshift(G);

%zero matrix L with a circle with value 1's with radius 50 
[xGrid,yGrid] = meshgrid(1:400,1:400);
L = sqrt((xGrid - 200).^2 + (yGrid - 200).^2) <= 50;

U = SAG2 .* L;

u = real(ifft2(ifftshift(U)));

figure
imshow(u,[]);