The next generation in digital video

The next generation in digital video

High-definition TV, film recordings on DVD and video clips on a cell phone – none of this would be possible without advanced image compression techniques. The successor to MPEG-2 is due to be launched in 2005, accompanied by better quality and a reduced volume of data.

The average German citizen switches on his TV for about three and a half hours per day – and the tendency is rising. The introduction of digital TV will increase the choice of programs for home viewing. DVB – Digital Video Broadcasting – will be based on a new standard that enhances the quality of transmitted images without requiring substantially greater bandwidth. The successor to MPEG-2 is called H.264. A team of researchers at the Fraunhofer Institute for Telecommunications HHI has made an important contribution to the performance of the new video compression standard. For this work, Thomas Wiegand, Detlev Marpe and Heiko Schwarz have been awarded one of the 2004 Joseph von Fraunhofer Prizes.

The compression program only re-saves those parts of each image that change from frame to frame. All parts of the picture that remain static in successive frames can be “predicted” from the previous frame. “Our contribution was to develop a prediction method that doesn’t merely rely on one previous frame, but uses a series of preceding frames,” as Wiegand explains. “It’s a more efficient way of predicting picture content that appears again after having been shown earlier. It does require more memory – but that’s a cheap commodity these days.”

Compression can be enhanced still further by raising the accuracy with which the image, divided into blocks, can be predicted. “We have succeeded in incorporating this increased accuracy into the specifications for the standard. Whereas the minimum size of the prediction blocks is 16-by-16 pixels in MPEG-2, it has been reduced to 4-by-4 pixels in H.264,” reports Wiegand. “This allows much more accurate predictions to be made.” The non-static (unpredictable) parts of the image are then quantized. This process takes out unimportant fine details of the image content that are not visible anyway – picture quality is in no way compromised. The now heavily compressed image content is then coded using a novel entropy method developed by the researchers. Rather like Morse code, frequently recurring image content is encoded in fewer bits than more seldom image content.

H.264 reduces the volume of data to somewhere between a half and a third compared to MPEG-2, depending on the application (digital TV, cell phone display or DVD recording). The DVD Forum has chosen H.264 as the new standard for all high-definition film records on this medium. The next generation in digital video will make its debut when High Definition DVDs start to become available next summer.
Thanks to Thomas.