Real Time Video Image Compression
Project Description
We are currently developing a technology that will make it easier for unmanned military and commercial vehicles to operate in hazardous areas. On a multiple processor video compression system provided by the U.S. Army, We apply computer vision and artificial intelligence techniques to develop real time video compression algorithms with high compression rates. We are developing mathematical models that reduce a video image's band-width to the degree required for visual transmission. The system is expected for use aboard an unmanned, remote-controlled vehicle. As the vehicle's video camera scans the hazardous area, the system compresses the images and transmits the compressed images to a remote location where a remote vehicle operator can view the images, in real-time, reconstructed images on a monitor. The operator can issue appropriate commands to control the movement of the vehicle based on the images she sees
There is a growing need for autonomous and remote-control vehicle systems for reconaissance, mine detection and the cleaning of hazardous waste. The compression procedure works similarly to the human eyes elementary stimulus for vision, perceiving complex patterns based on sparse contrast edges of an object being viewed. The widespread use of cartoons, sketches, blueprints and typefonts reflects this human talent.
The current system includes a 20-inch monitor, six parallel processors and two x-terminals. It employes technologies that reduced the band-width of video images transmitted from a remote vehicle. Conventional high-band-width video requires line-of-sight microwave or fiber-optic communications. Both types of communication pose problems. Microwave communication can expose the vehicle's position, while fiber-optic links restrict the vehicle's range and are prone to breakage.
The compression has three different stages. In the first stage, the system reduced the resolution of the image by digitizing the incoming video signal, splitting it into color and contrast channels. The procedure for color-channel compression mimics the way the brain processes images. Images relayed from the retina to the brain that are high resolution in the image's center change smoothly to low resolution in the periphery of the image. In the contrast channel, there is a compression procedure that models the human's visual capacity. The procedure works by extracting and transmitting encoded contrast edges. After completing these stages, the peripheral resolution of the video image is reduced to more closely match human perception, and large amount of information is reduced in transmission. The goal of the research project is to achieve transmission rate at 16kbps and keep the reconstructed images in good quality. To be effective, the display imagery was centered on the operator's gaze so that the image=92s high-resolution locus would follow the operator's eye. Objects in the scene will be defined and prioritized, and high-priority objects will hold their high resolution during compression. This research team is working on algorithms to prioritized objects in the scene. example, if a tank appeared at the edge of the video camera's sight, it would be given priority and be in focus. In addition to improving unmanned vehicle operations, this technology can be applied to teleconferencing, multimedia classroom technology and distance learning.