Optimization of Face Retrieval and Real Time Face Recognition Systems Using Heuristic Indexing

Dattatray D. Sawat, Kc Santosh, Ravindra S. HegadiRTIP2R 2020 · OpenReview

Abstract

Retrieving fast and accurate information is the key factor in real time biometric systems. The indexed data provides fast access to the pool of closest matching patterns in the database or the input given to the real-time systems. This paper aims to demonstrate the impact of heuristic indexing in realtime face recognition and on face retrieval systems using face skin color and complexion. This paper also presents the comparison of face recognition techniques with and without heuristic indexing and their impact on realtime performance. Proposed method works on both the indexed and non-indexed database. The survey shows that indexing not only improves retrieval speed but also rejects the wrong samples to be inspected, speeding up the overall system. We have proposed a novel heuristic indexing algorithm which locates key pixels on a face, calculates Skin Complexion Value (SCV) and their tolerance range, reduces search space and finally performs classification. The proposed approach yields significant improvement in terms of speed and reduction in False Acceptance Ratio (FAR).
BibTeX
@inproceedings{sawat2020optimization,
  author = {Dattatray D. Sawat and KC Santosh and Ravindra S. Hegadi},
  title = {{Optimization of Face Retrieval and Real Time Face Recognition Systems Using Heuristic Indexing}},
  booktitle = {RTIP2R},
  year = {2020},
  url = {https://openreview.net/forum?id=QRpk311MDL},
}

Loading related content...

Figure