Super Resolution
Super resolution is a process for producing a high resolution, high quality image from one or more low resolution images. This process is done either by inversing the imaging model or by hallucinating the high resolution details using some training database. Currently, I am working on this field and I am quite happy with it :)
Face Log Generation
Face logs can be considered as a cosine representation of a video sequence. Face analysis system working with video sequences can use the face log instead of the video sequence itself. Face logs can be generated for different purposes and by different methods, e.g. face logs used for indexing video sequence are generated differently than face logs for face registration or face super resolution.
Face Quality Assessment
Real world video sequences of face images contain lots of useless information. For example if the subject of interest is not facing the camera, if the subject is moving fast and his/her face is very blurry, if the face is too dark or too bright, if face size is too small, then such face images only add computational costs to the system. Therefore, applying any face analysis system to a video sequence of face images, without having some mechanism for discarding useless images would be extremely time consuming and likely to produce unstable response, e.g. in face image registration in video sequences. Face quality assessment is a very light and real time mechanism for discarding such images.
Face Detection
Even though there are plenty of methods for face detection, it is still a challenging problem. This is due to the different head pose, different facial expression, and different conditions of the real world environments. This problem would be more difficult for low resolution video sequences like in surveillance cameras.