Detection of Face and Objects on Eyes Boundary Using Color Model With Image Processing

Abstract

Finding facial features using image processing techniques is an important step in designing a facial recognition system. However, face detection still poses a challenge for research since the shape and appearance of each person’s face is different. There are also facial expressions, race, skin color,
and other factors including wearing objects on the face such as wearing glasses and hats that result in facial differences. The present study presents face area detection and eye object detection using the YCbCr color model and HSV model. The experiment used 200 images, divided into (1) 100 images
with no patterned background and (2) 100 images with patterned background. The YCbCr color modeling technique was applied to all images to distinguish the skin tone from the background. Because from previous studies, it was found that by YCbCr method, Cb and Cr color data values were similar to that of skin color spots covering all ethnic colors. HSV color model, which can clearly show the purity and brightness of the colors. Then the Sobel Edge Detection technique was used to detect faces. Next, an image segmentation technique was used to detect the area around the eyes to find the obscuring around the eyes. The results showed that for the background without pattern, 96% of faces were located and 91% of objects around the eyes were found, and for the patterned background, face location was 86% and subjects around the eye area were 82%, compared to face area detection with YCbCr color model alone. The results showed that the above method can increase the accuracy of 7% in detecting objects in the eye area of images with the non-patterned background. In the case of images with the patterned background, this method of locating faces can increase accuracy by 14% and increase the accuracy of finding objects around the eyes by 44%.