CSE473/573 Fall 2010 Lecture Charts

Chapter 3: Data structures for image analysis


 Matrices
 

The most common data structures for images are 2D

arrays, ie. matrices, whose pixel values represent

brightness information. For an n row by m column

image, M: {1,..,n}x{1,...,m} -> V.
 

    V=B={0,1}: binary image

    V={0,...,2b-1}: b bit intensity image
 

In case of  a color image, three matrices may be

used corresponding to RGB, CMY, YIQ, HSI, or

some other color model (Matlab refers to any

color image as an RGB image).
 


 

Indexed Image


Alternatively, an image can be described by

a combination of a look-up table (LUT) and a

matrix of values pointing to it. This is called

an indexed image.

For a color image, each LUT entry is set of three

numbers, the color values associated with the entry.

There may be any number of distinguishable colors

in the LUT.
 

The LUT is also called a color map CM, and the

combination of the LUT and the matrix of index values

is the indexed image.

    V={0,...,2b-1}: b bit indexed image

    CM: V -> C, a set of distinguishable colors.
 

Eg: In an 8-bit color image, pixel value V=23 may

be indexed to the color C=(255,0,0) or bright red.
 

An 8-bit indexed image has a pallette of 256

colors, a 24-bit image can map 224 = 16M colors.



Eg: Consider the quadtree representation of an 8x8 binary

noisy image of the letter H shown below. Each leaf of the

quadtree contains three kinds of information: the location

of the pixel set it corresponds to, the number of pixels,

and the color (black or white).