The object recognition based on Scale-Invariant feature transform and hybrid segmentation

Author

Zachariasova, Martina ; Hudec, Robert ; Benco, Miroslav ; Kamencay, Patrik

Author_Institution

Dept. of Telecommun. & Multimedia, Univ. of Zilina, Zilina, Slovakia

fYear

2012

fDate

21-22 May 2012

Firstpage

109

Lastpage

113

Abstract

This paper deals with research in the area of image analysis. Our approach is based on hybrid segmentation and Scale-Invariant Feature Transform (SIFT) method. The main idea is to improve the process of object recognition and their classification into classes by Support Vector Machine (SVM) classifier. The fast and powerful hybrid segmentation algorithm based on Mean Shift and Believe Propagation principles is used to improve object classification. Finally, the image segmentation algorithm was integrated with SIFT descriptor. The developed method was tested on real unsegmented and segmented images.

Keywords

feature extraction; image classification; image segmentation; object recognition; support vector machines; believe propagation principles; hybrid segmentation; image analysis; image segmentation; mean shift principles; object classification; object recognition; scale-invariant feature transform; support vector machine classifier; unsegmented images; Algorithm design and analysis; Belief propagation; Databases; Image segmentation; Kernel; Object recognition; Training; Belief Propagation; Mean Shift; SIFT; Semantic describe; segmentation;

fLanguage

English

Publisher

ieee

Conference_Titel

ELEKTRO, 2012

Conference_Location

Rajeck Teplice

Print_ISBN

978-1-4673-1180-9

Type

conf

DOI

10.1109/ELEKTRO.2012.6225582

Filename

6225582