Documentation for MDS/StructuredCollection

Technique StructuredCollection DS
Document MPEG-7 Final Committee Draft MDS, see StructuredCollection DS
Name Ana Belen Benitez, Columbia University
EMail ana@ee.columbia.edu
Type Application
External Libraries None
Related Ds/DSs Collection DSs, Model DSs, and ClusterModel DS
Used Ds/DSs ColorHistogram D
Input The input is a text file that describes the classification of M images
into N classes with the following format:
[Number of classes]
[Class name 1 without spaces]
[Number of images in the class] [image 1] [image 2] ...
[Class name 2 without spaces] [Number of images in the class] [image 1] [image 2] ...
...
[Class name N without spaces] [Number of images in the class] [image 1] [image 2] ...
An example of a well formed input file with four classes follows:
4
Grass 3 cultu~44_add2.jpg cultu~45_add2.jpg cultu~46_add2.jpg
Basketball 2 game111_add2.jpg game1191_add2.jpg
Person 3 i0132_add5.jpg i0122_add5.jpg i8h_add1.jpg
Dog 1 i8m_add1.jpg
Extraction Yes
Client Appl Search & Retrieval
Summary This code generates collection descriptions with associated probability
models for each class (sub-collection) in the input and the entire
collection of M images, and relations among the collections.
The probability models associated with a collection
consist of the mean and variance color histogram of the images in the
collection. This code also allows to retrieve collection structure descriptions but
matching based the closes two collections based on their mean color histogram.
Strong Points -
Limitations None
Parameters None