2D filter increases saved file size
1 view (last 30 days)
Show older comments
I have a question regarding spatial filtering... In the following (greatly) downsized example,
a = y0(1:100,1:2,1);
a = reshape(a,10,10,2);
r = zeros(size(a));
F = fspecial('gaussian');
for i = 1:2
r(:,:,i) = filter2(F,a(:,:,i));
end
save('r.mat','r')
save('a.mat','a')
the saved file a.mat has size 1Kb; and r.mat (2Kb). What leads to the increase in file size? (both 'a' and 'r' has the same dim and uses the same wrkspace memory).
Thanks!
0 Comments
Accepted Answer
Image Analyst
on 2 Nov 2011
I can pretty much guarantee that it doesn't take any more time to load a 1 kb file than a 2 kb file. Your disk reading software has a cache size of at least 8 KB of memory (<http://en.wikipedia.org/wiki/Disk_buffer>) so it sucks up a whole sector on disk in one shot. Anything in that sector gets read so it doesn't matter if it's 1 byte or 8000 bytes - it all takes the same amount of time.
But it looks to me like y0 is probably a uint8 color image. Am I right? So that would be 1 byte per element, while r would be a double image with 8 bytes per element. You think that might have something to do with it?
More Answers (3)
Image Analyst
on 2 Nov 2011
If you think about it you'll realize why the filtered image is bigger. The filter will "touch" the edge of the image even when the center of the filter is outside the bounds of the image. If you want filter2() to crop it to the same size, you can pass it the 'same' option.
2 Comments
Jan
on 2 Nov 2011
@bb: If you have a question concerning the command filter2, "help filter2" and "doc filter2" are very recommended. The documentation is written to solve such questions.
Jan
on 2 Nov 2011
Does a have the type double?
MAT files are compressed, if they are saved in the v7.3 style. Then the compression of a.mat has been more efficient than the compression of r.mat. But this shouldn't affect your work in any way.
2 Comments
Bjorn Gustavsson
on 2 Nov 2011
Then this puzzles me a bit. In my case filtered images have less noise and compress better. Are your image a synthetic, or even binary with a sparse number of non-zero pixels? If so, then the filtered image might have more non-zero pixels.
If the loading (and saving) of your filtered images takes longer time than it woudl take to load and filter the original image then you might as well just not save the processed images, but rather preprocess them on the fly as needed.
0 Comments
See Also
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!