|
|
|
|
| Representation by pixels |
Objects can be represented by sampling the sensor that observed them: pixels of an image, time samples of a time signal, frequencies of a spectrum, bins of a histogram. Although these are not all images we refer to this representation for all these types of objects by pixel representation. It can only result into a proper and useful vector space if for every object the same number of pixels are found and if objects are aligned. For images this means that they should have the same size (in pixels) and the objects in the images should have the same orientation and be positioned similarly.
Objects can be given the same pixel size by the Matlab routine imresize . The PRTools routine im_resize is a wrapper around imresize that operates on all images in a datafile or a dataset. The below routines may normalize images in various ways.
| > Preprocessing for Pixel Representation | |
| im_box | Bounding box |
| im_norm | Normalize images w.r.t. mean and variance |
| im_resize | Resize images |
| im_rotate | Rotate images |
| im_scale | Scale images |
| im_select_blob* | Select largest blob |
*The DIPimage package should be in the path for this command.
R.P.W. Duin
, January 28, 2013|
|
|
|
| Representation by pixels |