About Video Quality and Efficient Coding

Before we explore what efficient coding means, it is important first to note that all mainstream digital video distribution ses “lossy” compression, meaning that the encoded material when decoded will not match the original content exactly. We will use encoding and compression interchangeably in this discussion to mean a particular implementation of how high bitrate raw YUV digital moving pictures gets transformed into lower bitrate compressed digital information (such as MPEG2).

The emphasis here is on each actual implementation, which is a combination of algorithmic and processing limitations of each encoding system (or IC).

In the diagram below, we examine a particular 720x480 scene from two encoders encoded at 2Mbps

As can be seen from the example above, efficient encoding (delivering high quality at a low bitrate) is a concern because the choice of encoder can make a significant difference in how the final product is perceived by the end-user. Because the public is used to DVD quality video as a baseline, video recording products with visually or obviously inferior video will be an impediment to market penetration.

A comparative difference between the encoded content and the original content is one way of judging the quality of a compressor’s implementation, but this difference is not simply one of mathematical metrics. In particular, Peak Signal Noise Ratio (PSNR) analysis is a reasonable way to measure the accuracy of a particular compression, but often the best encoder is one that visually can analyze each picture in a series of moving pictures and determine what information to keep and what to discard.

For background information only, PSNR of a particular MxN bitmap Dcompared to the original bitmap S is taken as the following :

The equation is given for completeness, but we will not explore PSNR any further here.

In some cases, an encoding technique might discard information that is not visually noticeable to the human eye and this can cause lower PSNR -however, based upon viewing quality, this would still be judged to be the better encoder. Except under controlled environments with known input, in real world scenarios, arithmetic accuracy of lossy compression is not a measure of the quality of an encoder. This is a useful tool, but overreliance on this with random images can result in wrong choices when designing your encoder.

It is this problem that makes judging compression quality to be not simply one of mechanically applying simple arithmetic to frame by frame analysis of bitmaps.

上一页  [1] [2] [3] [4] [5] [6] [7] [8] 下一页