Lecture 5-Digital Video Compression and Sampling

Question 1

Define digitisation:

Answer 1

The process of converting an analogue signal to digital data.

Question 2

Define sampling:

Answer 2

A means of approximating an analogue signal digitally.

Question 3

define sampling rate:

Answer 3

Refers to how many samples are taken in one second.

Measured in cycles per second (Hz or MHz)

Question 4

What are the two formulas to measure signals at specific points of time for luminance component (Y) and Chrominance components?

Answer 4

R-Y or CR and B-Y or CB
Expressed as the ratio of samples per component Y:CR:CB
The number of samples per line equals the horizontal resolution.

Question 5

What is the base sampling rate of digital video?

Answer 5

13.5MHz

Question 6

What does Bit Depth refer to?

Answer 6

The level of detail in the data.

Question 7

What are the key components in the quality of digital video for 8-bit and 10-bit?

Answer 7

8-bit: 2^8 (or 256) levels of each colour component - R, G, B
16.7 million total colours in full palette (256R x 256G x 256B)
10-bit: 2^10 (or 1024) levels of each colour component - R, G, B
1.07 billion total colours in full palette
High end systems use 12 or 16 bits

Question 8

What is subsampling for 4:4:4?

Answer 8

For every 4 samples of Y there are 4 samples of CR and 4 samples of CB- 4:4:4
Highest quality sampling, for high-end transcoding, particularly for feature film, used extensively for visual effects work.

Question 9

What is subsampling for 4:2:2?

Answer 9

For every 4 samples of Y there are 2 samples of CR and 2 of CB. Very common high quality sampling. Used for broadcast television extensively. Acceptable but not optimal for visual effects.

Question 10

What is subsampling for 4:1:1?

Answer 10

For every 4 samples of Y, there is 1 sample of CR and CB.
Medium quality sampling, used for broadcast television for news or documentary, not used for visual effects work, more effective at higher resolution.

Question 11

What is subsampling at 4:2:0?

Answer 11

For every 4 samples of Y there are 2 samples of CR and 2 samples of CB on alternate lines. CR and CB are Interpolated for lines without chroma samples. Medium quality sampling. Similar applications to 4:1:1 more effective at higher resolution.

Question 12

Define quantization:

Answer 12

The reducing of luminance or Chrominance information into a bit depth.

Question 13

Define Banding and Dithering:

Answer 13

Banding: bands of colour become visible when the bit depth does not allow enough colours to accurately show gradient of colour.
Dithering: is a means of mixing adjacent shades to make banding less apparent.

Question 14

Define aliasing and anti-aliasing:

Answer 14

Aliasing: there is insufficient chrominance information to show clean transitions between areas of significant colour difference. Sometimes called stair stepping, noticeable on curved or diagonal edges of objects in frame.
Anti-aliasing: similar to dithering in that it uses different shades to reduce visible aliasing.

Question 15

Define compression:

How is it measured?

Answer 15

Compression is encoding data using less detailed data than the original.
Data is measured in bytes: 8 bits=1 byte
Compression is different to sampling
CODEC=colour-decoder
Codecs process sequences of frames for compression

Question 16

Define lossless and lossy:

Answer 16

Lossless: compression reduces the amount of storage (or bandwidth) required but all data is intact.
Lossy: compression can reduce the amount of storage (or bandwidth) required but some data is loss.
Different applications require different types of compression.

Question 17

In lossless compression, what is run-length encoding (RLE)?

Answer 17

A means of aggregating like areas of data designated by number of pixels, R value, G value, B value. For example, if a scan line has 7 samples of pure red followed by 6 samples of pure black, that section could be encoded as 7, 255, 0, 0, 6, 0, 0, 0: this uses 8 bytes instead of 39 bytes (13x3).

Question 18

In lossless compression what is LZW?

Answer 18

LZW is a more sophisticated type of lossless compression. Makes a table noting all the values it comes across. Then stores the notes.

Question 19

In Lossless compression what is TIFF, DPX, OpenEXR and RAW?

Answer 19

TIFF is the most common format, still image scheme but used for sequences in visual effects.
DPX is widely used in visual effects industry, has advantage of incorporating metadata (i.e. Additional data about lens settings, camera settings, other data).
OpenEXR is a new format for visual effects.
Note that Raw is not a CODEC nor a standard format and refers to unprocessed data captured by the imager.

Question 20

What is the alpha channel?

Answer 20

A fourth channel stored in some file formats e.g. TIFF, sometimes described as RGBA/ARGB, not usually captured by cameras.
Usually created in VFX software. Used extensively in VFX for compositing. Can be used to store extra information about the image: transparency etc.

Question 21

Explain some features of lossy compression:

Answer 21

Many types of it. Data is identified that is not significant to the quality of the image and then discarded. Adaptive compression systems adjust parameters automatically. The less data kept, the smaller the file but the poorer the image. Types of visual degradation vary depending on CODEC e.g. MPEG gets blocky, Wavelet loses sharpness etc.

Question 22

Define intraframe coding:

Answer 22

Each frame encoded individually. Chroma is subsampled- (either 4:1:1 or 4:2:0).
Audio is uncompressed- 16 bit 48kHz PCM stereo or four tracks of 12 bit 32kHz PCM.

Question 23

In lossy compression- MPEG, define interframe, spatial and temporal.

Answer 23

Interframe coding uses spatial and temporal compression.
Spatial compression uses an analysis of frequencies within groups of pixels, stores an array of coefficients.
Temporal compression looks for repeated data between different frames, stores an instance and then discards it. Only changes of data within and between frames are kept.

Question 24

In lossy compression what is the standard MPEG for SD and HD broadcast television?

Answer 24

SD MPEG-2
HD MPEG-4
AVC Inta and H.264 are variants of MPEG-4
Note that some formats, like Sony’s XDCAM are simply wrappers for different compression schemes.
XDCAM, QuickTime, AVI and MXF can all use DV, MPEG-2 or MPEG-4 based on compression schemes as well as others.

Question 25

What are the calculations for storage requirements?

Answer 25

Storage per frame (bytes)=resolution x 3 x bit depth / 8
Note there are 3 colours and 8 bits in 1 byte
Storage per second=storage per frame x frame rate (multiply by 60 for minutes, by 60 again for hours). Result gives the total requirement for uncompressed video.
To calculate the effect of compression, divide by ratio
Total storage = uncompressed storage/compression ratio.

Question 26

5 minutes of SD PAL video, uncompressed?

Answer 26

720 x 575 x 3 x 8 bits / 8 bits per byte
=1,244,160 bytes per frame (or 1.9MB)
1,19 MB x 25fps x 60 sec/min = 8.7 GB total
Remember there are 1,024 bytes in 1 kilobyte (KB),
1,024 KB in 1 megabyte (MB)
1,024 MB in 1 gigabyte (GB)
1,024 GB in 1 terabyte (TB) etc.

Question 27

5 hours of HDTV 24p video (4:2:2) using HDCAM SR?

Answer 27

HD resolution is 1920 x 1080 pixels
HDCAM SR is a 10 bit format
1920 x 1080 x 3 x 10 / 8 = 7.41 MB per frame
7.41 x 24 x 60 x 5 = 3.06 TB total uncompressed
2.06 / 2.7= 1.14 TB total