Strange 3x3 Lowpass filter

I need to perform an article about hdr imagining and i see a formula that smooth the pixel.But it is different formulla. Can someone help me to code it on matlab

but i think first iteration will be problem. when i and j start with 1 and k is -1 first iteration will be ymax(0,0) but there is no such a thing like that in MATLAB.I think we should add +1 near i+k and j+l.

Thank you

Answer 1

[Chad Greene]

It looks like a pretty straightforward 3x3 averaging filter. The catch is you have to start with (i=2,j=2) so i+(-1) will be 1 and same for j. You can do this pixel by pixel, or you can do it much faster with imfilter:

ylpf = imfilter(ymax,fspecial('average',[3 3]));

No loops necessary.

[Chad Greene]

If you do not have the Image Processing toolbox, you can use conv2 instead:

P = peaks(50) + 2*rand(50);

% With image toolbox: 
Pf = imfilter(P,fspecial('average',[3 3]));

% Without image toolbox: 
Pf2 = conv2(P,ones(3)/9,'same');

figure
subplot(1,3,1) 
imagesc(P) 
title 'noisy data'

subplot(1,3,2) 
imagesc(Pf) 
title 'imfilter' 

subplot(1,3,3) 
imagesc(Pf2)
title 'conv2'

Answer 2

[Walter Roberson]

There is no problem if you restrict i to be in [2, (number of rows - 1)] and restrict j to be in [2, (number of columns - 1)]

For areas outside that, the formula is not well defined.

Note: consider using the single statement

y_lpf = conv2(y_max, ones(3,3)/9, 'same');

... except, that is, that the result is well defined at the edges.