Today, market with one of the highest growth and development is computer and mobile industry market, and in last few years there is rapid interest in devices that have the possibility of total immersion into a virtual world. Although this technology is usually linked with video games, platform provides new possibilites for video, e-advertising and many more. One of the possibilities that this technology offers is also virtual chatting, where the person we are talking to can be animated character as desired. That is not currently possible, because half of the face is covered with HMD. Although that problem can be easily overcomed, we still have one problem to solve. Namely, none of todays softwares for facial tracking have the ability to track only lower part of human face. This paper bypasses stated problems by assuming one camera mounted on HMD that records only lower part of human face, and then it joins that picture with predefined image of human face ensuring that given overlay looks as natural as possible. Before any frame is processed, camera is calibrated with assumption of radial and tangential distortion. After that, every frame is undistorted by coefficients calculated in calibration process. Afterwards, histogram of every frame is modelled towards histogram of background so the contrast and brightness would be more similar to the background. Finally, processed image is merged into a whole using binary mask that is filtered with Gaussian kernel. Given solution is independent of specific HMD or special sort of camera.