How can I resize a rectangular image of size 1280x512 to a square image of size (2048x2048)? I have used the following code. Is it correct or is there any other better methods to perform resizing?

% First read the image I of size (1280x512) using imread, then used imresize to change the size of I to (2048x2048) and used imwrite to save the image

I=imread('I.tif');
J=imresize(I,[2048 2048],'bicubic');
imwrite(J,'J.tif','tiff');

Answer 1

[Walter Roberson]

It depends on your purpose. If the purpose is the human visual system, then you might possibly get more-appreciated output using imresize() with 'lanczos3'. See http://en.wikipedia.org/wiki/Lanczos_resampling#Limitations

[VENKATESAN]

Thanks for your answer.. But, still, how to use this lanczos3 kernel in the code? Whether changing a rectangular image to a square image will change the field of view of the image?

[Walter Roberson]

J=imresize(I,[2048 2048],'lanczos3');

I am not sure what you mean by "field of view" in this case? No new information will be introduced in the resizing. The aspect ratio will be changed so that circles in the original image will become ovals that are about twice as wide as they are tall. There is no way around changing the aspect ratio of objects in the image when you change the aspect ratio through resizing. You could, though, resize keeping the same aspect ratio, and then pad part of it to take you up to a square image (that had no actual content for about 1/4 of the image.)

[VENKATESAN]

Thank you very much for your details about image resizing. Actually, the resolution of our image depends on field of view (FOV). Resolution = FOV/No. of pixels. FOV defines the distance of the total number of pixels in an image in millimeter, so that we can able to calculate the size of the object that we are imaging. Since we are changing the dimension of the image from (1280x512) to (2048x2048), whether the change in number of pixels will affect FOV.

[Image Analyst]

Gee, Hollywood & Burbank do it all the time. In movies and TV, you just zoom into your crummy surveillance video by a factor of 50 or 100 and now you can read license plates and newspapers and see peoples' faces in super high resolution. But my favorite scene was in the movie "No way out" http://www.youtube.com/watch?v=m4SBMyd0yEQ where they had a Polaroid negative of Kevin Costner and the CIA or FBI was spending days trying to process this thing, which was getting more and more focused pass after pass. Of course all they had to do was to negate the image, but anyway... Kevin Costner wanted to delay this process so he got his friend in charge of the group to delay it and he went by the image analysts and told them "The eigenvalue is off. We're pulling away from our reference information. Program a Fourier transform." Friends wondered why I laugh out loud at that scene.

When you change the size of your image, of course it does not change the field of view. Increasing the resolution in pixels will not increase the real-world spatial resolution in mm either. However decreasing the size of your image will reduce the size of objects you can resolve in your image (because you're throwing away information).