Efficient small face detection in surveillance images using major color component and LDA scheme

  • Authors:

  • Kyunghwan Baek;Heejun Jang;Youngjun Han;Hernsoo Hahn

  • Affiliations:

  • School of Electronic Engineering, Soongsil University, Dongjak-ku, Seoul, Korea;School of Electronic Engineering, Soongsil University, Dongjak-ku, Seoul, Korea;School of Electronic Engineering, Soongsil University, Dongjak-ku, Seoul, Korea;School of Electronic Engineering, Soongsil University, Dongjak-ku, Seoul, Korea

  • Venue:
  • CIS'05 Proceedings of the 2005 international conference on Computational Intelligence and Security - Volume Part II
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

Since the surveillance cameras are usually covering wide area, human faces appear quite small. Therefore, their features are not identifiable and their skin color varies easily even under the stationary lighting conditions so that the face region cannot be detected. To emphasize the detection of small faces in the surveillance systems, this paper proposes the three stage algorithm: a head region is estimated by the DCFR(Detection of Candidate for Face Regions) scheme in the first stage, the face region is searched inside the head region using the MCC(major color component) in the second stage and its faceness is tested by the LDA(Linear Discriminant Analysis) scheme in the third stage. The MCC scheme detects the face region using the features of a face region with reference to the brightness and the lighting environment and the LDA scheme considers the statistical features of a global face region. The experimental results have shown that the proposed algorithm shows the performance superior to the other methods’ in detection of small faces.