Facial landmark detection using FAST Corner Detector of UGC-DDMC Face Database of Tripura tribes

Abstract

We have performed facial landmark detection of visible face images and studied the performance of Features from Accelerated Segment Test (FAST) Corner Detector technique on IRIS Face Database and newly created UGC-DDMC Face Database to justify the design approach of the face database. This paper describes detection of the facial landmarks of UGC-DDMC face database which includes left eye corners, right eye corners, left eyebrows, right eyebrows, lip corners, nostrils. It consists of two parts: In first part, a morphological opening operation is used to estimate the background. Then create a more uniform background by subtract the background image from original image. In the second part, the facial landmarks had been detected using Fast corner detector technique. The Fast corner detector works on the corner response function (CRF), which is computed as a minimum change of intensity over all possible direction. The Fast corner detector is significantly faster to compute than other algorithms. The experiment has been done over the UGC-DDMC Face Database of Tribes of, which is being developed by IT Department of Dasaratha Deb Memorial College, Khowai, Tripura. In our Experiment initially we used 108 Facial images with different pose from the UGC-DDMC Face Database and 110 images from IRIS Database. Performance results for ntIS database is 100% and UGC-DDMC Database is 74%. It has been seen that algorithmic performance is high with low threshold value and with higher values if shows good performance. Fiducial points obtained during this process will be useful for any feature based Face recognition process. Thus, we can justify the design process of the UGC-DDMC Face Database.