28.02.2013 Views

Bio-medical Ontologies Maintenance and Change Management

Bio-medical Ontologies Maintenance and Change Management

Bio-medical Ontologies Maintenance and Change Management

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3-D color histogram<br />

M<br />

Tc Qc -1 M -1<br />

T (Qc ) c<br />

Multimedia Medical Databases 97<br />

Fig. 3.4. Transformation between 3-D color histogram <strong>and</strong> binary color set<br />

University of Columbia [83, 84, 85, 87]. The binary color sets can be obtained<br />

by selection from M colors. The process is represented in the following figure<br />

[84, 86]:<br />

Let B M be an M dimensional binary space, such that each index value m<br />

corresponds to one axis in B M .<br />

The following definition can be given [86]:<br />

The color set s c is a binary vector in B M . A color set corresponds to a selection<br />

of colors from the quantized color space.<br />

A binary set is equivalent with a limited histogram. For example, having the<br />

threshold τ m for a color m, the color set is obtained using the following equation:<br />

Sc [m] = 1, if hc [m] ≥ τm , otherwise Sc [m] = 0 (3.14)<br />

The color set specifies only the colors that are above the specified threshold.<br />

The color sets give a good representation of colors, images <strong>and</strong> regions. If a<br />

color is not representative in a region <strong>and</strong> it is under the threshold τm it is ignored.<br />

In this manner, with the color set, the color content is represented using only the<br />

most preeminent colors from image or region.<br />

3.5.3 Color Quantization<br />

Because the color spaces are multi-dimensional, a partitioning of the color space is<br />

described by a spece vector quantization [84, 85]. Generally speaking, a vector<br />

quantization Qc of dimension k <strong>and</strong> size M is a mapping of a vector from a kdimensional<br />

space, in a finite set C which has M outputs. Qc can be defined as a<br />

function Qc : ℜ k<br />

→C, where C = (y0 , y1 , …, yM-1 ) <strong>and</strong> ym ∈ ℜ k<br />

for each m ∈ 0,1, …,<br />

M-1 [84, 86].<br />

The set C is called the collection of cods <strong>and</strong> has the size M. For the case of<br />

vector quantization of a color space, k = 3 <strong>and</strong> for each input in the codes<br />

collection, the y corresponds to a color point.<br />

m<br />

There is a coverage of ℜ k with M partitions that is associated with the vector<br />

quantization, where each partition R has all the points w assigned to the same<br />

m c<br />

code y : m<br />

1<br />

B M<br />

1<br />

Binary color sets<br />

1

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!