Supervised and Unsupervised Data Engineering for Multimedia Data

2Unsupervised/Supervised Feature Extraction and Feature Selection for Multimedia Data (Feature extraction with feature selection for Image Forgery Detection)

Arun Anoop M.^1*, Karthikeyan P.² and S. Poonkuntran³

¹Alva’s Institute of Engineering & Technology, Mijar, Moodbidri, Karnataka, India

²Velammal College of Engineering and Technology, Viraganoor, Madurai, Tamilnadu, India

³VIT Bhopal University, Sehore, Bhopal, Madhya Pradesh, India

Abstract

Multimedia data needs to be protected from unauthorized duplication, otherwise, data may lead to tampering which may not be identified by the naked eye. Features and feature vector are collections of local and global features of digital images. Features can determine the image level classification accurately nowadays. So, the importance of feature extraction is high. Some redundant and irrelevant data may be many in the case of feature extraction. To avoid it, feature scaling and feature selection approaches are mandatory to get an accurate prediction. Multimedia data are mostly image, video and audio technologies. In this paper, we demonstrate different supervised and unsupervised feature extraction algorithms and forgery classification technique of 42 features to check the accuracy of detection and supervised image classification based on GLCM(24), GLDM(4) and GLRLM(7) feature extraction combination with LBP(7) variants. Later the same method will process, based on a different correlation map, a threshold-based feature selection ...