Добавил:
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
литература / Digital_Video_and_HD_Second_Edition_Algorithms_and_Interfaces.pdf
Скачиваний:
0
Добавлен:
13.05.2026
Размер:
38.02 Mб
Скачать

A rule of thumb that relates data rate to storage capacity:

Mb/s = GB/movie

Gb/s = TB/movie

Introduction to

video compression

16

Directly storing or transmitting digital video requires fairly high data capacity – about 20 megabytes per second for SD, or about 120 megabytes per second for HD. Here is a rule of thumb that relates storage capacity and data rate: Eight, 2000-ft reels of motion picture print film can carry a 1331/3 minute movie; there are 8 bits in a byte and 60 seconds in a minute; and

60/8 ·1331/3 is 1000. So one megabit per second equals one gigabyte per movie – whether compressed or not! Similarly, one gigabit per second equals one terabyte per movie.

Economical storage or transmission requires compression. This chapter introduces the JPEG, M-JPEG, MPEG, and H.264 compression techniques.

In previous chapters, we have discussed representation of image data in a rather small number of colour components (say, three); a rather small number of bits per component (say 8 or 10); perceptual coding by way of a nonlinear EOCF; and chroma subsampling yielding a data rate reduction of around 50%. In video terminology, all of these techniques are termed paradoxically, perhaps to be uncompressed video. Compression involves transform techniques such as the discrete cosine transform (DCT) and – in the case of

JPEG 2000 – the discrete wavelet transform (DWT).

Data compression

Data compression has the goal of reducing the number of bits required to store or convey text, numeric, binary, image, sound, or other data. High performance is obtained by exploiting statistical properties of the data.

147

Salomon, David (2008), A Concise Introduction to Data Compression

(Springer).

Sayood, Khalid (2005), Introduction to Data Compression, Third edition(Elsevier/Morgan-Kaufmann).

The term “perceptually lossless” signifies an attempt to minimize the perceptibility of compression errors. There are no standards or industry practices to determine to what extent that goal is achieved. Thus, the term is indistinct.

The reduction comes at the expense of some computational effort to compress and decompress. Data compression is, by definition, lossless: Decompression recovers exactly, bit for bit (or byte for byte), the data that was presented to the compressor.

Binary data typical of general computer applications often has patterns of repeating byte strings. Most data compression techniques, including run-length encoding (RLE) and Lempel-Ziv-Welch (LZW), accomplish compression by taking advantage of repeated strings; performance is highly dependent upon the data being compressed.

Image compression

Image data typically has strong vertical, horizontal, and spatial correlations among samples of the same colour component. When the RLE and LZW algorithms are applied to bilevel or pseudocolour image data stored in scan-line order, horizontal correlation among pixels can be exploited to some degree; such techniques usually result in modest compression (perhaps 2:1).

A data compression algorithm can be designed to exploit the statistics of image data, as opposed to arbitrary binary data; improved compression is then possible. For example, the ITU-T fax standard for bilevel image data exploits vertical and horizontal correlation to achieve typical compression ratios higher than RLE or LZW typically achieve. In the absence of channel errors, data compression (even of images) is lossless, by definition: Decompression reproduces, bit-for-bit, the data presented to the compressor.

Lossy compression

Lossless data compression can be optimized to achieve modest compression of continuous-tone (greyscale or truecolour) image data. However, if exact reconstruction is not required, the characteristics of human perception can be exploited to achieve dramatically higher compression ratios: Image or sound data can be subject to lossy compression, provided that any impairments introduced are not overly perceptible. Lossy compression techniques are not appropriate for bilevel or pseudocolour images; however, they are very effective for greyscale or truecolour images, both stills and video.

148

DIGITAL VIDEO AND HD ALGORITHMS AND INTERFACES

 

Uncompressed

 

Compression ratio

 

Format

Motion-JPEG

MPEG-2

H.264

data rate [MB/s]

 

 

 

 

 

SD

20

15:1

45:1

90:1

(480i30, 576i25)

 

(e.g., DV25)

(e.g,. DVD)

 

 

 

 

 

 

HD

120

20:1

75:1

100:1

(720p60, 1080i30)

 

 

(e.g., ATSC)

(e.g., Blu-ray)

 

 

 

 

 

Table 16.1 Approximate compression ratios for SD and HD video distribution systems

Internet protocol television (IPTV) concerns video and audio delivered over TCP/IP networks.

Encoders and decoders in compression systems are not to be confused with composite video (NTSC or PAL) encoders or decoders.

JPEG stands for Joint Photographic Experts Group, constituted by ISO and IEC in collaboration with ITU-T (the former CCITT).

Transform techniques are effective for compression of continuous-tone (greyscale or truecolour) image data. The discrete cosine transform (DCT) has been developed and optimized over the last few decades; it is the method of choice for continuous-tone image compression. JPEG refers to a lossy compression method for still images. MPEG refers to a lossy compression standard for video sequences; MPEG-2 is used in digital television distribution (e.g., ATSC and DVB), and in DVD. H.264 refers to a lossy compression standard for video sequences. H.264 is highly effective for HD; it is used in satellite, cable, and telco (IPTV) systems, and in Blu-ray. These techniques will all be described in subsequent sections.

Table 16.1 compares typical compression ratios of M-JPEG and MPEG-2, for SD and HD.

In the context of compression of video or audio, the term codec refers to an enCOder and/or a DECoder.

JPEG

In 1992, the JPEG committee adopted a standard based upon DCT transform coding, suitable for compressing greyscale or truecolour still images. This was before the world-wide web: The standard was expected to be used for colour fax! JPEG was quickly adopted and widely deployed for still images in desktop graphics and digital photography. The M-JPEG variant can be used for motion sequences; the DV scheme uses an M-JPEG-like algorithm. Details are presented in JPEG and motion-JPEG (M-JPEG) compression, on page 491.

A JPEG compressor ordinarily transforms R’G’B’ to Y’CBCR, then applies 4:2:0 chroma subsampling to effect 2:1 compression prior to the transform coding steps. (In desktop graphics, this 2:1 factor is included in the compression ratio.) JPEG has provisions to compress R’G’B’ data directly, without subsampling.

CHAPTER 16

INTRODUCTION TO VIDEO COMPRESSION

149

Compression ratio

Quality/application

Example SD tape formats

 

 

 

2:1

“Visually lossless”

Digital Betacam

 

studio video

 

 

 

 

3.3:1

Excellent-quality studio video

DVCPRO50, D-9 (Digital-S)

 

 

 

6.6:1

Good-quality studio video;

D-7 (DVCPRO), DVCAM, consumer DV

 

consumer digital video

 

 

 

 

Table 16.2 Approximate compression ratios of M-JPEG for SD applications

JPEG and motion-JPEG (M-JPEG) compression is described on page 491. DV compression is described on page 505.

Taubman, David S. and Marcellin,

Michael W. (2002), JPEG-2000:

Image Compression Fundamentals,

Standards and Practice (Norwell,

Mass.: Kluwer).

Motion-JPEG

The JPEG algorithm – though not the ISO/IEC JPEG standard – has been adapted to compress motion video. Motion-JPEG simply compresses each field or frame of a video sequence as a self-contained compressed picture – each field or frame is intra-coded. Because pictures are compressed individually, an M-JPEG video sequence can be easily edited; however, no advantage is taken of temporal coherence.

Video data is almost always presented to an M-JPEG compression system in Y’CBCR subsampled form. (In video, the 2:1 factor due to chroma subsampling is generally not included in the compression ratio.)

The M-JPEG technique achieves compression ratios ranging from about 2:1 to about 20:1. The 20 MB/s data rate of SD can be compressed to about 20 Mb/s, suitable for recording on consumer digital videotape (e.g., DVC). M-JPEG compression ratios and tape formats are summarized in Table 16.2.

JPEG 2000

Between 1995 and 2000, the JPEG committee developed a compression standard for continuous-tone colour still images. The effort culminated in the JPEG 2000 standard, which is based upon discrete wavelet transform (DWT) techniques. DCI standards for digital cinema use JPEG 2000 compression. An adaptation of JPEG 2000 accommodates motion sequences, where each (progressive) frame is coded individually without reference to any other frame. Although the “core” JPEG 2000 coding system is intended to be royalty and license-free, intellectual property rights (IPR) concerns have inhibited JPEG 2000 commercialization.

150

DIGITAL VIDEO AND HD ALGORITHMS AND INTERFACES

Соседние файлы в папке литература