face detection result analysis..

The application developed in MATLAB for image processing based attendance marking was tested both manually and automatically. First database creation was done using the photos of students in the university. Experiments were carried out using 22 photos. Among them, the face recognition
algorithm was able to recognize 15 students correctly. Hence the percentage error obtained was 31.81. The eigen faces calculated are displayed as shown in fig. 9 and the attendance report generated for the students in the university is shown in
Fig. 10.


Usually in the eigen face method, faces are cropped manually to remove the background [2]. But in this proposed system, the objective is to recognize the individuals from a photo automatically. So it was first needed to detect and extract the faces from the photo by using some face detection algorithm and then these extracted faces were to be used for the
recognition. As the size of the face extracted by the face detection algorithm was small, there was a high percentage error as above mentioned was experienced in recognizing individuals. Also, there are some more limitations for the eigen face approach. First, the algorithm is sensitive to head
scale [8]. Second, it is applicable only to front views. Third, it demonstrates good performance only under controlled background, and may fail in natural scenes. When lighting is highly variable, eigenface often does no better than random guessing would [1, 7, 2]. Other factors that may stretch image