Author Topic: Vision Inspired Local Ternary Pattern (VLTP) for Face Recognition & Verification  (Read 36 times)

Offline iftekhar.swe

  • Full Member
  • ***
  • Posts: 144
  • মানুষ তার স্বপ্নের সমান বড়
    • View Profile
    • DIU_SWE Faculty
Abstract: Face recognition and verification algorithms use a varity of features that describe a face. Most popular amongst these features are LBP (Local binary pattern) and its varient Local Ternary Pattern (LTP). LBP is very sensitive to near uniform region and is incapable of handling intensity fluctuation that often happens due to noise. This is addressed by introducing a fixed threshold in LTP. However, a fixed threshold often fails to perfectly describe a feature. To address this issue, we propose an adaptive LTP (ALTP) that extends LTP to evoke vibrant threshold. To verify the proposed methods we have used a recent challenging face database named Label Face in Wild (LFW). Our proposed ALTP method is light weight, and achieved an accuracy of 76.23%, which is impressive in contrast to other computationally inexpensive state of the art methods.

Sr. Lecturer
Department of Software Engineering, FSIT
Daffodil International Univeristy