|
contents | software | |||||||
| Fotonation Announces Face Tracker for Camera Phones FotoNation has announced FotoNation Face Tracker for camera phones. Face Tracker uses a radically new approach to identify and lock onto human faces in a camera phone's preview image, tracking them as they move around within the frame and automatically adjusting focus, exposure, and white balance before the image is captured, ensuring that faces are optimally taken and that skin tones are reproduced with exceptional accuracy. Especially important for shooting under poor lighting conditions and at night-time, Face Tracker draws a box around the face, showing the user which face is being tracked. It can also track multiple faces in the preview image simultaneously. Because of its advanced tracking algorithm, Face Tracker accurately follows and adjusts exposure on the selected faces as they move in frontal or profile positions and establishes the correct image orientation automatically. FotoNation Face Tracker also takes into account any processing speed limitations within the camera phone. The Face Tracker technology is highly optimized, allowing sufficient processing power for the camera phone to run its own algorithms while faces are being detected, continuously maximizing the available resources. Available for digital cameras as well as camera phones, Face Tracker enables the novice user to make high quality portrait pictures effortlessly. Until now, this has been the specialist domain of professional photographers using high-end cameras. Face Tracker Features: - In-camera phone real-time face detection and tracking - Quickly works on down-sampled image - Excellent general-purpose detection - Scene-dependent parameters are collected and used to improve accuracy - Tracks faces in a wide range of poses - Once face is "locked, " momentary pose changes do not effect processing of image - Optimized detection - Carefully designed to allow other simultaneous processing in the camera phone. - First detection run detects faces throughout the entire image - Once a face is detected, detection takes place only in areas where a face was seen before - The technology uses prediction to anticipate face movement - When large changes are detected, the full frame is searched for new faces - Automatic Image Orientation - High accuracy: 96% correct rotation if a dominant face is present - A voting algorithm decides the orientation when multiple faces are present in the image - The correct rotation decision is based on a sequence of preview frames so the accuracy is much better than a decision based on a single image approach - The orientation algorithm can deal with profiles in final image if frontal profile has been present (because of tracking) - Orientation can be stored in Exif tag - Very simple API for easy integration into the camera phone. write your comments about the article :: © 2006 Computing News :: home page |