Implementation Of Glcm And KNN Algorithms For Classification Of Real Faces And Screen Displays

Abstract

The significance of image processing technology has grown tremendously in recent years, with its applications now encompassing various fields, including biometric sistems. A salient concern in the realm of biometric sistems pertains to the potential for face forgery through images exhibited on mobile device screens. The objective of this research is to develop a method for detecting skin texture that can differentiate between real human skin and skin displayed on a mobile device's interface. The proposed approach utilizes the Gray Level Co-occurrence Matrix (GLCM) to analyze the skin texture, while the K-Nearest Neighbor (KNN) algorithm is employed to differentiate between human skin and the screen. The GLCM operates through the extraction of texture features, including contrast, energy, homogeneity, and correlation, from grayscale images. These features are then trained and tested using the KNN algorithm to classify the image into the real skin or screen category. Then LBP (Local Binary Pattern) was added to see the micro texture which can help GLCM detect texture more accurately. The dataset under consideration in this study is comprised of a series of images depicting human skin, along with images of human skin displayed on a mobile phone screen. These images have undergone a series of processing steps, including preprocessing and feature normalization. The evaluation results demonstrate a classification accuracy rate of 90%, suggesting that the method is effective in identifying distinctive texture patterns between human skin and screen. This research also contributes to the development of biometric sistems, such as face-based attendance, which are designed to detect fraud, such as the use of fake faces from the screen.