how to detect blurred images in a dataset.

I have one thousand frames converted from a video taken from a particular location using gopro camera. From that dataset, I need to detect the blurred images and segregate that alone, how to do that???

Answer 1

[Image Analyst]

Perhaps you can take the fft of the image and check if the energy in the higher frequencies is not very high.

[Vimal Raj]

I'm Getting the fft like this, how to check the frequency level.

the code i used is this,

clear all;

close all;

clc

I=imread('E:\ak\6.jpg');

figure (1); imshow(I); title('Input Image');

I=rgb2gray(I);

figure (2); imshow(I); title('gray Image');

F=fft2(I);

s=abs(F);

figure (3); imshow(s,[]); title('FFT of Image');

Fsh=fftshift(F);

figure (4); imshow(abs(Fsh),[]); title('centered fft');

s2=log(1+abs(Fsh));

figure (5); imshow(s2,[]); title('log transformed Image');

% F=ifftshift(Fsh);

% f=ifft2(F);

% figure (6); imshow(f,[]); title('reconstruted Image');

[Image Analyst]

Just sum up the energy outside some box or circle like I said. Something like

[rows, columns] = size(s2)
mask = true(rows, columns);
row1 = round(0.4*rows)
row2 = round(0.6*rows)
col1 = round(0.4*columns)
col2 = round(0.6*columns)
mask(row1:row2, col1:col2) = false; % Make a hole in the middle of the mask image so we won't count that area.
sumOfEnergy = sum(s2(mask))

[Vimal Raj]

Thank you sir, but could you please share some reference material or can you describe elaborately with code.

[Image Analyst]

Just do a web search on "Fourier filtering image" and I'm sure you'll find lots of stuff to look at.

[Vimal Raj]

Ok sir, thank you.