Facial expression recognition is an interesting and challenging problem, and found in many applications like humancomputer interaction (HCI), robotics, video surveillance, border security, clinical research, person verification, crime prevention etc.. Facial expression is the movement of the muscles beneath the skin of the face. Through facial expressions human can convey their emotions without any verbal means. In this paper we have created raw database of color images. Training and testing set of images are created. Color information in an image is used to detect the face from the image. Important features from the detected face are extracted to form feature vectors using Gabor and Log Gabor filters. Principal Component Analysis (PCA) is used to reduce the dimension of the extracted features. Then these reduced features are classified using Euclidean distance. The main aim is to work upon three emotions-happy, neutral, surprise. Experiment carried out on self-generated database show comparable performance between Gabor and Log Gabor filters, where Log Gabor filters outperforming Gabor filters with classification accuracy of 86.7%.