|
Consumers are ready for wireless video. Sleek ads commonly show a new HD LCD TV hanging unobtrusively on a wall, with no hint of associated wires or cables. We know that there must be a power outlet in the wall directly behind it, but are the AV cables also built into the wall? In the early days, when flat panel televisions were only for high-end homes, the cables may very well have been pulled in behind the wall. Today, though, HD flat panel televisions are rapidly reaching mainstream adoption, and the average homeowner wants the same clean look shown in the ads, without having to go through the pain of home remodeling. And, with the plethora of AV connections possible—HDMI, component, USB, VGA—a clean wireless solution has tremendous appeal.
Benefits of Video Compression
High-bandwidth wireless solutions enable wireless video, but aren’t the whole answer. Today, compression is still required because of bandwidth limitations, and errors in the channel will increase the actual bandwidth required.
An uncompressed 720p video stream consumes 1.485 Gbps; this doesn’t include bandwidth required for audio, encryption, and packetization. Wireless vendors strive to increase the bandwidth and quality of the wireless channel, and it is arguable that uncompressed video will be feasible within the next several years. There are two main benefits to sending uncompressed video: low latency and high error resiliency. Since there is no compression, the only delay will be the delay of the channel and the processing that already exists in wired solutions. Uncompressed video is also highly resilient to bit errors; the human eye will integrate out random errors, and in extremely noisy conditions the large number of random errors will manifest as static.
Although uncompressed wireless video provides the lowest latency and best error resiliency, video compression offers two significant advantages. First, compressed video consumes much less bandwidth, allowing the channel to be used in other ways. The channel can be shared among other applications, multiple video streams could be transmitted simultaneously, or the range of the channel could be increased (by allowing more bandwidth to be used for retransmissions). Second, in the case of truly scalable video codecs, the same compressed video stream can be sent to multiple displays of various resolutions, with each display decoding only the portions that it requires. Uncompressed video received by a display with a different resolution than the source must be scaled in the display, adding significant cost, latency and complexity to handheld devices.
Challenges of Wireless Video
Wireless video holds three major challenges: error resiliency, latency, and image quality in the face of variable channel bandwidth.
Low latency is especially important for applications requiring real-time video compression. Gaming applications immediately come to mind, but playing a DVD is also a real-time application. It likely won’t be acceptable to wait 2-3 seconds after pressing the fast-forward button before seeing the video respond. DVD menu commands, to play the director’s cut, for instance, also must occur with low latency.
Menu overlays added in the video source (DVD player or set-top box) require that the source content, say a DVD, be uncompressed in the box before the menus are overlaid. Adding menu overlay in the video source has several benefits—STB makers get to keep the coveted user interface, no new specifications for menu side-channel transmission are required, and latency when a remote control button is pushed can be minimized—but it requires the output to be compressed in real-time or that it not be compressed at all before wireless transmission. For upcoming HD-DVD and Blu-Ray discs, the real-time compression must also be HD capable. While it is possible to send menu command information over a side channel and send the pre-compressed DVD content wirelessly, there are a couple of disadvantages to this approach. The TV must perform the rendering functions; graphics rendering chips are updated at a much faster rate than TVs, so rendering in the TV adds cost but not value, as this is a function consumers already expect to be performed by the DVD player. Also, HD-DVD and Blu-Ray pre-compressed content can have a highly variable output bit rate – between 30 Mbps and 40 Mbps – which will require sophisticated processing to be added to the TV.
Wireless channels typically perform forward error correction (FEC) and packet retransmissions, decreasing visible errors but increasing complexity of the receive side by requiring potentially very large buffer memories. In addition, the latency increases and becomes variable, since retransmissions are variable with error rate. Obviously, reduction of retransmissions decreases memory requirements and decreases latency. Reduction of FEC also increases available bandwidth, which could be used to increase range or increase the number of AV channels supplied. Different codecs manifest errors in different ways; a codec that is highly resilient to errors requires less FEC and retransmissions in order to produce acceptable video.
Image quality is the final consideration for video compression schemes. Ideally, the compression process will not produce any artifacts in the reconstructed stream: this is called lossless compression. If the artifacts are not visible to the Human Visual System (HVS), then the compression is call visually lossless. Compression is the control of output bit rate and compression efficiency is often referred to in terms of the ratio of the compressed to the uncompressed video bit rate (e.g., 10:1 or 20:1). As long as the input video to the compression engines is the same (identical filtering and decimation), and as long as the resulting video quality is subjectively similar, compression ratios provide a quick and useful way to compare the efficiency of various codecs. Image quality and efficiency are different for error-free (wired) and noisy (wireless) channels; the same compression ratio cannot be used for both unless the wireless channel has no errors. The ability of a codec to change the output bit rate as the channel degrades is also a serious challenge. If the output bit rate is 35 Mbps and the channel can support only 30Mbps, the codec must respond or the image may not be reproduced at all.
JPEG2000 for Wireless Video
With these challenges in mind, the characteristics of JPEG2000 for wireless video applications are considered in the rest of this article.
JPEG2000 is a scalable, wavelet-based video-compression standard ratified as ISO/IEC 15444-2 and ITU-T Rec. T.800 in 2004. Compression is performed on an intra-frame basis only, using a wavelet compression method. The use of wavelets produces a hierarchical codestream that can be scaled in terms of quality and resolution, which will be discussed later. The hierarchical codestream also creates high innate error resiliency; a review of the codestream generation is required to understand why.
Compression can be thought of as bit-rate control. JPEG2000 has two distinct mechanisms for rate control: the wavelet transformation process and entropy encoding.
Wavelet Transform
High pass and low pass filters are applied to the rows and columns of each frame, producing four subbands.
This process is repeated on the “low low”, or LL subband (the upper left corner). In figure 1, two image resolutions can be recovered: all four subbands combine to recreate the original resolution, and the LL subband is a 1/4 size resolution all on its own. (Images courtesy of Michael W. Marcellin and Ali Bilgin – University of Arizona)
Figure 1: One 2D Transform (2 resolutions)
The image below shows two transform levels; five or six levels are typically used.
Figure 2: Two 2D Transform Levels (3 resolutions)
The LL subband is the most important for recreating the image, while other subbands provide additional detail and resolution. Subbands are further divided into quality levels. The wavelet transform very lightly quantizes all the frequency components in each transform level and each subband, resulting in 2 to 3 compression. The codestream packets can be ordered in many different ways; the method below shows progression by resolution.
Figure 3: Wavelet Transform Illustration
Progression by resolution allows resolution scalability; a device does not have to decode the entire stream to reproduce a smaller resolution image. A handheld device, for instance, could decode only the portions of the codestream necessary for its small display, while the same stream could be fully decoded by an HDTV.
Next: Entropy Encoding
|
