Matlab imfilter uint8 overflow

Hello,

I created a gaussian filter. Then I divide every element by the smallest element so my gaussian filter is in integer. When i convolve an image with it, I get 255 for everything. How do I fix this ? I want to convolve using integer values not decimals. Thanks.

Answer 1

[Walter Roberson]

double() the data before convolving.

[RuiQi]

I am trying to prove the linearity of the gaussian ie G(Ia + Ib) = G(Ia) + G(Ib). I tried subtracting the result before getting the max element in the result G(Ia+Ib) - G(Ia) - G(Ib) but I cant seem to get a zero.

[Walter Roberson]

That is a different matter. You asked how to avoid saturating at 255, and that question has been answered.

[RuiQi]

ok but imshow gives me white what do i do

[Guillaume]

With doubles, by default imshow assumes the range [0 1] for intensities. Convert the data back to uint8, or specify the DisplayRange in imshow to be [0 255]:

imshow(uint8(yourimage));
%or
imshow(yourimage, [0 255]);

[Walter Roberson]

imagesc(yourimage)

to display it automatically scaled according to the data it contains. This will not be the same as either of Guillaume's suggestions (which would normally be valid) because your filters are producing values larger than 255.