Digital video compression methods
Gradually, all the equipment for the production of television and video programs in the world is going digital. Why is this happening?
It is not at all due to the fact that everyone admires digital methods and that the image quality is radically improving, the reasons for such a transition are mainly economic, because now both the price and the cost of operating a digital complex are lower. than a traditional analog with the same functions. Now that digital technology has reached a certain maturity, analog equipment is suddenly less efficient, less reliable and much less profitable.
EN The very foundations of video production and delivery of images to the viewer are currently being revised, and the dynamics of change in the current age – digital – is much higher than in analog. Today, for example, the 4: 3 aspect ratio is gradually being replaced by the 16: 9 aspect ratio, interlaced-progressive (non-interlaced) scanning and changing encoding standards increase image quality. The economy also dictates the need to compress video information. Naturally, any compression degrades the quality of the rendering and is applied not for a good life, but out of necessity.
Do not think that video compression was not used before and that it only appeared with the advent of digital technologies. No, compression has always been used, but only before it was analog, and today it is necessary to make it completely digital, getting rid, as much as possible, of double compression, that is, both analog and digital.
Modern digital compression methods, which replaced analog, especially when combined with computer technology, can not only improve the quality of the video itself, but also expand the possibilities of video production and optimize the display of audiovisual products.
Basic concepts of digital transformation
Essentially, a digital representation, or digitization, is a partition of the domain of a continuous function in some intervals and the representation of this function as a set of values at the end of these intervals. So, for a digital audio signal, the second interval on the time scale is divided into 32, 44 or 48 intervals, in each of these samples the sound is measured and its value is stored with a certain representation precision, generally 14 to 20 bits. These operations are called sampling processes, that is, the representation of any continuous quantity (in this case, sound) through periodic discrete measurements. After that, they say that the sound is digitized with a sample rate of 32 to 48 kHz, respectively, and a bit depth of 14 to 20 bits. Therefore, the digital current
In digital video, the sampling process for a value is simply generalized to a multivalued function that has a range of values, not a number, but an image (that is, it becomes three-dimensional) to break solid lines of a value. image (scanning beam) in so-called pixels, or image elements, our three-dimensional space will be because when the image is presented at each moment of time, not a value is involved, as in sound, but the whole image – a frame . As a result, the video signal is also represented as a sequence of integers, in other words, as a stream of data.
In the case of (separate) component analog video signals, each channel must simultaneously transmit three color components (red – red, green – green and blue – blue, i.e. RGB) or a luminance signal together with signals of color difference (YUV). The conversion of RGB color components to YUV components is one by one, but the latter method makes it possible to highlight the change in signal brightness, for which the human eye provides a separate perception mechanism, with the help of rod cells. in the retina, in contrast to the perception of color by cone cells. Rods are more sensitive than cones, so in systems with separate component coding (e.g.
Sampling theory, in turn, requires that the sampling frequency be at least twice the bandwidth of the signal to be digitized. In the case of a broadband signal (for sound, that is, for example, the simultaneous reproduction of several octaves of frequencies), the sampling frequency must be at least twice the highest frequency of the input signal analog (Fig. 1).
Figure: 1. An example of sampling an analog signal: a – correct; b – wrong. In the first case, the original signal is easily reconstructed in the form of a sample envelope, and in the second, when the signal frequency exceeds half the frequency
