Real-Life Images

This paper presents a detailed idea about the face detection technique which reduces false positive rate and improves the rate of detection. Face detection remains as one of the challenging problems in the field of image analysis and computer vision. The color image is converted into a gray-scale image. Based on the spatial localities and orientation characteristics of the Gabor filter, we extract the facial feature from the facial image by appropriate filter design. The input of feed-forward neural network classifier is the feature vector obtained by the Gabor filter. The convolution is done by multiplying the image with the Gabor filter in the frequency domain and is saved in a cell array in order to save time. Thus the input of the network will have larger values and is optimized by reducing the matrix size to one-third by deleting appropriate number of rows and columns. The results obtained in the output layer in testing with real-life images shows that the false-positive rate is reduced by 97%. Index Terms:Gabor filter, Feed-Forward Neural Network, computer vision, color image processing