Texture classification by pattern knowledge discovery

Shi, Hui (2007) Texture classification by pattern knowledge discovery. Masters thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (3MB)


Texture analysis has received a considerable amount of attention over the last few decades as it creates the basis of the most object recognition methods. Texture analysis mainly comprises texture classification, texture segmentation, and both of them require the important step: texture features extraction. Many approaches have been proposed either as spatial domain methods or frequency domain methods. Many texture features based on the spatial domain methods have been proposed as those methods are proven to be more superior. Texture can also be considered as a collection of patterns. Distances, directions and pixel gray-level values can determine the relationship among pixels within each pattern. Therefore, patterns are considered as the basis of textures and textures are considered to be different if they contain distinguished patterns. The procedure of pattern knowledge discovery has been started in order to find the distinctive texture patterns with gray-level deviations and distances deviations. An apriori algorithm with the joining step, cleaning step and pruning step has been introduced to find frequent patterns in order to generate higher order patterns which can be used to categorize textures. A large number of textures from the benchmark album of Brodatz have been applied and tested in the proposed method in order to prove the validity of this system and the performance is promising. The overall high accuracy shows the great encouragement from the testing procedure.

Item Type: Thesis (Masters)
URI: http://research.library.mun.ca/id/eprint/11424
Item ID: 11424
Additional Information: Includes bibliographical references (leaves 62-66).
Department(s): Science, Faculty of > Computer Science
Date: 2007
Date Type: Submission
Library of Congress Subject Heading: Computer algorithms; Pattern recognition systems.

Actions (login required)

View Item View Item


Downloads per month over the past year

View more statistics