I Face Detection And Recognition System author: Mukund Agarwal supervisor: Professor Nishan Canagarajah Project Thesis submitted in support of the Degree of Bachelor of Engineering in Electronic and Communications Engineering

7 
3 . 2 I m p l e m e n t a t i o n
The implementation is done using the software package MATLAB®. The motion detection 
module is closely based on the motion detection graphical user interface developed by David 
Tarkowski
4
. The functions used are taken from the image acquisition and processing toolboxes 
provided in MATLAB®. 
A video input object is created by using the ‘videoinput’ function this provides a 
communication bridge between MATLAB® and the video capture device. The adapter which 
enables this communication is ‘winvideo’ for this system. 
The implemented code uses ‘imabsdiff’ function (1) to calculate the absolute difference 
between two images. This function utilises Intel Performance Primitives Library (IPPL) which 
accelerates its execution time. IPPL contains basic functions used in image and signal 
processing. 
imabsdiff (frame new ,frame background) = frame new - frame background ... (1) 
graythresh{imabsdiff(frame new ,frame background)} = result where 0≤ result ≥1 ... (2) 
Figure 4 Difference image and its binary image 
The resultant difference image is then passed to the ‘graythresh’ function (2). This function 
uses Otsu’s method
5
to reduce the graylevel image to a binary image (figure 4) and then 
calculate the optimum threshold which ensures that intra class variance is minimal.

---

Mukund Agarwal 
Face Detection & Recognition System 
8 
So if the passed image’s resolution is 1280x1024 pixels and is in RGB mode, then it will go 
through each of the (1280x1024x3) = 3932160 cells and will process all of them to give one 
value. This value will be normalized. If there is motion present then the difference image will 
contain some other values then black {RGB (0, 0, 0)} in those cells. If nothing changed between 
those two consecutive frames then this means that all of the cells will be zero. In summary this 
function processes each RGB value and summarises the difference in one value. (Figure 5)