The Performance of LTS-based Regression Methods in Face Recognition with Occluded Images

Abstract

In this paper, we compared the performance of several Least Trimmed Squares (LTS) based methods for face image recognition. The focus was on the problem of severely occluded face images. The performance of random LTS, Fast LTS and GA-LTS (which is based on genetic algorithm) to a benchmark dataset with occluded query images was examined. The best method was the one being least affected by the occluded images and produces highest recognition rates. The AT&T and Yale Data were used to assess the methods in performing face recognition. The query images were contaminated with salt and pepper noise and the recognition rates were measured when the contaminated images were used as query image in the context of linear regression. Results show that the random LTS outperforms the rest in dealing with occluded images with higest recognition rates.