Image Motion Processing

With the increasing demand for consumer content, bandwidth is becoming a major constraint, whether for television or for video over the Internet. There simply aren't enough transponders to carry the projected requirements for television channels much beyond the next few years. And the problem can't be solved just by launching more broadcast satellites. Orbital slots at geosynchronous altitudes are at a premium, and the cost isn't going down.

For commercial television broadcasting, the solution has been a combination of technologies. There are more transponders carried on board each satellite, which helps somewhat to relieve congestion. However, the real solution is to compress the images being broadcast, so that each channel uses a minimum amount of bandwidth and thus allows more channels per transponder. Much work has been done to achieve these goals, and data compression standards such as MPEG-2 and MPEG-4 are being implemented by different manufacturers for use in broadcast systems. Similarly, over the Internet, compression of video sequences using MPEG-2 is common.

The key to successful compression of video imagery is understanding what happens from frame-to-frame within the image. If the inner motion is known, then picture elements (pixels) that don't move can be compressed efficiently by avoiding retransmission of redundant data. Similarly, projecting motion through multiple frames means that the moving pixels can be compressed by only passing information about the motion and the new locations of the data, rather than by frame-to-frame transmission of each pixel in each image. Compression approaches such as MPEG use motion detection algorithms to allow more compression that single-frame static image compression.

Martingale Research has developed an Image Motion Processing algorithm that can be used as a component of overall compression schemes such as MPEG or H-320 (video conferencing). The algorithm processes the entire image and provides velocity estimates at the pixel level for efficient motion compensation and compression.

Back to top

Key Features

  • Operates at the pixel-level - NOT a block-based algorithm.
  • Very efficient - requires few computations per pixel
  • Highly parallelizable - can be effficiently implemented into custom silicon
  • X and Y velocity outputs for each pixel in the image
  • Robust in the presence of noise and flicker
  • Also applicable to motion tracking applications
Back to top

Technical References

The Martingale Research Image Motion Processor algorithm is inherently designed to work with non-linear, time-varying information such as video imagery. Click here to view a set of slides discussing the technical underpinnings of our algorithm.

Back to top

Product Applications

Embedding Martingale's Image Motion Processor within compression / decompression systems or hardware will provide new capabilities for bandwidth reduction and increased utilization of restricted communications channels. Click here to view a set of slides discussing product applicability of the Image Motion Processor algorithm.

Back to top

Additional Information

Relative Performance
Advanced Motion Estimation Applications
Back to top Skip to navigation