Core Information Extraction Using Multilinear Manifold Learning and Hierarchical Algorithm for Image Understanding in Large Scale Dataset

2014 Fiscal Year Final Research Report

• Project/Area Number: 24700179
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Perception information processing/Intelligent robotics
• Research Institution: Ritsumeikan University
• Principal Investigator: HAN Xian-Hua 立命館大学, 立命館グローバル・イノベーション研究機構, 准教授 (60469195)
• Co-Investigator(Renkei-kenkyūsha): CHEN Yen-Wei 立命館大学, 情報理工学部, 教授 (60236841)
• Project Period (FY): 2012-04-01 – 2015-03-31
• Keywords: 画像認識 / 多次元多様体学習 / データ駆動モデル / 局所特徴 / テンソル解析

Outline of Final Research Achievements

In this project, we propose to represent an image as a local descriptor tensor and use a Multilinear Supervised Neighborhood Embedding (MSNE) for discriminant feature extraction, which is able to be used for subject or scene recognition. The contributions of our project include: (1) a image representation framework denoted as local descriptor tensor, which can effectively combine a moderate amount of local features together for image representation and be more efficient than the popular existing Bag-Of-Feature model; (2) a MSNE analysis algorithm, which can directly deal with the local descriptor tensor for extracting discriminant and compact features, and at the same time preserve neighborhood structure in tensor space; (3) a data-driven modeling procedure for raw features instead of the hand-craft local descriptors such as SIFT, SURF. We demonstrate the performance advantages of our proposed approach over existing techniques on understanding different types of benchmark database.

Free Research Field

機械学習、パターン認識、画像処理