Feature extraction of images texture based on co-occurrence matrix

Joint Authors

Abd Allah, Hadiyah S.
Yasin, Wijdan
al-Sayf, Khalil Ibrahim

Source

ZANCO Journal of Pure and Applied Sciences

Publisher

Salahaddin University-Erbil Department of Scientific Publications

Publication Date

2019-04-30

Country of Publication

Iraq

No. of Pages

10

Main Subjects

Social Sciences (Multidisciplinary)

English Abstract

There are many techniques to extracted object properties in an image.

In this research a co-occurrence matrix has been to adopted for feature extraction of English letters.

English letters of size 14 and font time new roman have been stored as image, then preproced by apply truncation to take off all blank area, then filtered to make it noise free.

Energy, contrast, correlation and homogeneity of the co-occurrence matrix properties for the stored character images were calculated.

Another character models with different size and fonts were adopted to make the database able to cover a wide range of character images for character recognition and classification .

Applied technique shows that a companation features can be extracted as new properties for letters images and give good results.

The experimental results of the proposed algorithm have proved that both energy and homogeneity features have given high recognition compared with the remaining other properties

Data Type

Conference Papers

Record ID

BIM-894469

• APA
• MLA
• AMA