Octave/Matlab高提升滤波实现：高斯低通模糊后锐化代码求助

Hey there! Let's wrap up your image processing pipeline and make sure that high-boost filtering sharpens that blurry moon just right. Your initial Gaussian low-pass implementation is solid—let's fix the loose ends and walk through the high-boost step in detail.

First, Fix the Low-Pass Output (Critical Step!) #

Your current InverseGauss is a complex array from the inverse FFT, so we need to extract the real component and normalize it back to the 0-255 range (and convert to uint8 for proper image display) before moving to high-boost filtering.

Complete Code with High-Boost Implementation #

Here's the full, polished version of your code with explanatory comments:

% Load the blurry input image
I = imread('blurry-moon.tif');
figure('Name','Image Processing Pipeline');

% Step 1: Gaussian Low-Pass Filter in Frequency Domain
A = fft2(double(I));
Ashift = fftshift(A);
[m, n] = size(A);
R = 10; % Adjust this radius for more/less smoothing
X = 0:n-1;
Y = 0:m-1;
[X, Y] = meshgrid(X, Y);
Cx = 0.5*n;
Cy = 0.5*m;
% Gaussian low-pass filter kernel
LoF = exp(-((X-Cx).^2 + (Y-Cy).^2) ./ (2*R^2));
% Apply filter and inverse transform
Gauss = Ashift .* LoF;
GaussShift = ifftshift(Gauss);
InverseGauss = ifft2(GaussShift);

% Clean up the low-pass result (remove complex artifacts, normalize)
I_lowpass = real(InverseGauss);
I_lowpass = uint8(mat2gray(I_lowpass)*255); % Convert back to 0-255 uint8

% Step 2: High-Boost Filtering
A_boost = 2; % Boost factor (1 = no sharpening, >1 = increasing sharpness)
% Convert images to double for safe arithmetic operations
I_double = double(I);
I_lowpass_double = double(I_lowpass);
% Calculate high-pass component (original - low-pass)
I_highpass = I_double - I_lowpass_double;
% Apply high-boost formula: J = I + (A-1)*I_highpass
J = I_double + (A_boost - 1)*I_highpass;

% Clamp values to 0-255 to avoid overflow/underflow, convert back to uint8
J = uint8(max(min(J, 255), 0));

% Display all stages for visual comparison
subplot(1,3,1); imshow(I); title('Original Blurry Image');
subplot(1,3,2); imshow(I_lowpass); title('Gaussian Low-Pass Filtered');
subplot(1,3,3); imshow(J); title(['High-Boost Sharpened (A=',num2str(A_boost),')']);

Key Technical Notes & Tuning Tips #

• Gaussian Radius (R): Increasing R will smooth the image more, which means the subsequent high-pass component will capture stronger edge details. Try values between 5 and 20 to find the sweet spot for your moon image.
• Boost Factor (A_boost):
  • A_boost = 1 returns the original image (no sharpening)
  • Values between 1.5 and 3 are ideal for subtle to moderate sharpening
  • Avoid values >4, as they'll amplify noise and create unnatural, over-sharpened edges—test incrementally!
• Data Type Safety: Always convert uint8 images to double before arithmetic operations to avoid integer overflow. The clamp step (max(min(J,255),0)) ensures your final image stays within valid pixel value ranges.
• No Ringing Artifacts: Unlike ideal low-pass filters, Gaussian low-pass filters don't cause ringing around edges, making them perfect for this sharpening workflow.

内容的提问来源于stack exchange，提问作者potu1304