Video segmentation & Retrieval

Abstract

There is a tremendous growth in the fields of multimedia and web databases, and research has been stepping forward towards many computer vision applications. In many computer vision applications local features are needed. To address this specific issue, many large point descriptors and detectors have been invented throughout the years. Creation of effective descriptors is still a milestone. To combat the high computational cost and the hunger for training data, auto encoders are proposed for efficient image analysis and image retrieval. Based on the auto encoder concept, a novel descriptor has been introduced. The proposed descriptor reduces the size and complexity and hence reduces the time required by a database to produce and display the retrieval results.