Image Analysis

Contents

Intensity Histogram - the image intensity distribution

Classification - mapping from pixel values to feature or object classes

Connected Components Labeling - grouping pixels with the same class label into regions

Many of the other categories of image processing operators demonstrated in HIPR transform an input image to form a new image without attempting to extract usable global information from the image. The operators included in this section do extract globally useful information, such as:

• Pixel Value Distribution, the number of pixels having each value.

• Classified Pixels, the category of scene entity that the corresponding pixel is taken from.

• Connected Components, the groups of pixels all of which have the same label or classification.

It is possible to represent some of this information as an image (i.e. an image whose pixels are the category or index number of the scene structure), but the information need not always be so represented, nor even necessarily be representable as an image. For example, while not included here, one might analyze a binary image of an isolated object to determine the area or various moments of the shape. Another example would be to link together connected edge fragments to make a list of edge pixels.

This category of operation is often considered part of the middle level image interpretation (i.e. a signal-to-symbol transformations or feature extraction), and whose results might ultimately be used in higher level image interpretation (i.e. symbol-to-symbol transformations such as scene description, object location, etc.).

©2003 R. Fisher, S. Perkins, A. Walker and E. Wolfart.