|
Research in the area of Multimedia Processing concentrates on issues related to the modeling of multimedia signals (still images, image sequences, video, music, speech, etc.) as well as their processing in order to facilitate information extraction, compression and transmission. Current research activity centers primarily around visual information. A wide range of interesting topics are investigated, with the main focus on multidimensional signal modeling, motion modeling, estimation and segmentation, video compression and transmission, watermarking and authentication, 3-D multimedia systems design. Applications that motivate this research are primarily encountered in video surveillance, video compression for various new services (Internet, cellular and wireless), new human-computer interfaces (3-D), next-generation multimedia services, etc.
Two main research thrusts are currently in place:
- Image Processing
Faculty: Bystrom, Karl, Konrad
Image processing is a rapidly evolving field with growing applications in both professional and consumer markets. While computer-aided tomography (CAT), extensively used in life sciences, is an example of application in the professional market, digital photography is an example of a very successful application in the consumer market. Fundamental issues studied in this thrust cover problems related to visual perception, sampling and quantization of visual data, image restoration and enhancement, image segmentation, and image recognition, as well as image transmission and storage (image compression, watermarking, authentication, etc.)
- Video and Image Sequence Processing
Faculty: Bystrom, Konrad
Image sequence is the next step, after a still image, in assuring a faithful capture of scene characteristics; in addition to spatial detail (texture, color) it captures dynamics occuring in the scene (e.g., motion of objects, change of illumination). Digital video is a special case of an image sequence, characterized by a particular sampling pattern and color representation, that has found a very prominent place in the entertainment market (DVD, miniDV, streaming video). In this thrust, advanced solutions are sought to such challenges as enhancement, segmentation (moving object extraction) and compression of image sequences. One particular video characteristic that is exploited is the temporal coherence of visual data along motion trajectories; various video processing and compression methods are being developed by jointly processing the visual data in spatial (horizontal and vertical) and temporal coordinates.
- 3-D Imaging
Faculty: Konrad
Another step in assuring a "life-like" visual experience is to exploit the binocular aspect of human vision, i.e., the fact that humans perceive depth based on the capture of environment from two slightly different perspectives. The capture of two images by closely-positioned cameras (about 2.5in apart) followed by their separate presentation to viewer's eyes, invokes the perception of depth and is known as "stereoscopic" imaging. When the two views presented to the left/right eyes are allowed to change with viewer head motion, and thus render other perspectives (the so-called "look-around"), terms "multiscopic" or "multiview" imaging are commonly used. In this thrust, we are concerned with problems ranging from data multiplexing for 3-D displays, through the generation of virtual views, to 3-D image compositing and multiview image sequence compression. Applications of this work can be found in entertainment, medicine and life sciences, tele-operation.
- Multimedia Transmission
Faculty: Bystrom, Konrad
Multimedia transmission is new and rapidly growing field which is concerned with all aspects of processing and manipulating multimedia data for transmission and storage. Fundamental issues in this area include data compression and coding, preprocessing (such as pre-filtering), interaction with physical transmission storage elements, and post-processing such as voice or video restoration. There are many applications which range from video-on-demand consumer services to ad hoc network design to medical data storage, transmission, and access.
- Auditory Signal Processing
Faculty: Nawab
Auditory Signal Processing is concerned with the the digital representation and manipulation of microphone data. The methods utilized may come from digital signal procesing, statistical signal processing, and/or the study of the human auditory system. Applications of interest include speech restoration, speech enhancement, speaker separation, auditory scene analysis, and automatic music transcription.
|
