Semi Supervised Video Object Mining Framework to Multiple Object Extraction

Abstract

Video mining using combination of supervised and unsupervised learning techniques has created an arc in multimedia data mining. With this binding (usual & unusual mining) technique, today we can get accurate results in multimedia applications. This blend takes away the formal techniques that were used in video mining. Though video mining in general it is purely unusual mining, there exists some complex computational work by all means. Hence this work first compares the efficiency in video mining between unsupervised and semi-supervised learning techniques and then proposing a model or framework for multiple object extraction. In Multimedia Mining multiple object extraction is one of the challenging areas. This is because it contains more critical issues and it is a complex task when it comes to dynamic applications. Hence an attempt is made with some assumptions to extract multiple objects using semi supervised learning techniques. This proposed model blends semi supervised learning techniques and multiple object extraction with necessary compression and decompression methods in a simple way as an initial step to address the two challenging areas of video mining.