Normalization is the process of converting a random sized image into a standard size. To bring all characters into a common size platform in order to extract features on the same footing, a minimum bounding box is fitted to the character and the element is cropped and then resized to fit into 32x64 window. In, the cropped element is normalized to a size of 36x36 pixels without disturbing the aspect ratio using bilinear standard transformation. There is another method for resampling which is based on equiarc length.
Dear Pritam.
Your question needs a bit of clarification, what are you exactly trying to do? What is a "random sized image"?
For converting images, you can download the ImageMagick package. It contains the "convert" program which can scale/crop/etc, images of any size into the size you like.
http://www.imagemagick.org/script/download.php
If you are using Matlab, then the function 'imresize' will do the job. For example if you have to resize image A to B while specifying the size, then:
B = imresize(A, [YY XX]);
where YY and XX are scalars representing final sizes in pixels.
Also, if you want to preserver aspect ratio of any dimension, just use NaN instead of giving scalar value in the above. For example:
B = imresize(A, [100 NaN]);
This will only resize one dimension (height) to 100 pixels while preserving the aspect ratio for the second dimension.
if you are using opencv, this site can be useful: http://docs.opencv.org/modules/imgproc/doc/geometric_transformations.html
Regards
- Badges
- Science topic
- Similar topics
- Computer Science
- Images