Extended Multiple Means Fuzzy Local Gravity Edge Operator for Efficient Feature Extraction Using Machine Learning

In many real-time applications, face recognition with noisy images is challenging task. The random variation of illumination or color information in images is known as image noise and noise may occur due to aging conditions like wrinkles, dark spots etc., Noise has been shown to have a great impact on the performance of face recognition methods. When an image consists of noise then the accuracy of face recognition methods would drop significantly. To improve the face recognition system accuracy rate in the presence of noise and rotation variant images, A novel edge detection technique using multiple mean local gravity edge operator is proposed. This method divides the image into eight sub regions and mean is calculated for each subregion separately then maximum of mean is subtracted from the original image finally, the gravitational force exerted by mean values of each subregion to the central pixel is computed using both horizontal and vertical directions. This Technique effectively removes noise from the face images but fails to recognise illumination variant face images.