Performance Comparison Of Out-Of-Plane Facial Detection Using Speeded Up Robust Features (SURF) And Scale Invariant Feature Transform (SIFT)

Nor Ain Zuzila, Zolkifly (2015) Performance Comparison Of Out-Of-Plane Facial Detection Using Speeded Up Robust Features (SURF) And Scale Invariant Feature Transform (SIFT). Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 pages) Performance Comparison Of Out-Of-Plane Facial Detection Using Speeded Up Robust Features (SURF) And Scale Invariant Feature Transform (SIFT) 24 Pages.pdf - Submitted Version Download (498kB)

Abstract

Nowadays, SURF and SIFT become a popular method for facial detection due to its advantages in detecting and recognizing of features. The aim of this research is to implement an algorithm for face detection using SURF and SIFT as well as to test an algorithm for face detection using SURF and SIFT. As SURF and SIFT has their own steps, the different algorithm is used to evaluate the performance of time and the number of feature point detection using SURF and SIFT. The research is started with implement the algorithm of both techniques. Then, the facial image has been captured by using camera with different pose that indicate the out-of plane rotation. After that, the facial image has been tested through MATLAB software. Finally, the performance comparison of out-of-plane facial detection using SURF and SIFT has been evaluated. In general, SURF and SIFT are scale invariant. SURF uses Haar wavelet and it is the fastest feature detector and descriptor. Meanwhile, SIFT uses Difference of Gaussian over various scales of an image and it is the most accurate feature detector and descriptor. It is expected that SURF is better in terms of speed, while SIFT is the best for accuracy. The parameters that have been evaluated from both techniques are time and the number of feature point detection. As a conclusion the performance of technique chosen between SURF and SIFT will be compared.

Actions (login required)

View Item