Many applications require segmented facial features by which a person can be recognized. Face recognition, Face expression recognition, Security system and wide variety of application needs facial features for proper identification. This paper describes locating facial features like eyes, eyebrows, and nose and mouth regions. And in turn uses those extracted features for recognition purpose. But it is quite difficult to extract features from noisy images. To segment the facial features from noisy image, a new model is developed. Thus, a novel system that works on noisy image without using any filtering algorithm is proposed. This work segments the facial features from noisy image which has salt & pepper noise of density value of 0.04. Moreover in order to improve the segmentation process of facial features, computing efficiency, and recognition rate further the Modified Active Contour (AC) algorithm is used. In this, the entire work divided into five different stages. At first, from the noisy query image, the contrast suppressed image is obtained. The second stage outlines the facial features of both query and target images by using Modified Active Contour (AC) method. Third stage extracts features from both the images. In the fourth stage, the height and width of the extracted features are measured for recognition. Lastly, comparison is performed based on the measured values. A match or a mismatch is the outcome of the system. The proposed system tested on different datasets and the experimental results on Chicago face database shows this work will drastically increase the segmentation accuracy of features compared to other existing systems but gives a substantial rate of recognition.
Volume 12 | 04-Special Issue