More Essays  | [Bill's Home]

 

 

Click on me, if you just want to see the magazine

Date: 2 August 2001
Ed: Billatnapier
Source: Chapter 8, Mastering Computing
Price: £10 (UK)/ $14 (US)
 
This is a special edition of the Compression Agent, and presents some of the most asked questions relating to data compression. More information can be found in Mastering Computing.
[Show only magazine][Next edition]
 
Question: You say that high-quality audio uses 16 bits for each sample, but my CD player says that it uses 1- bit conversion. Is this right?
Yes. It does. It uses one bit at a time, as this is thought to give a smoother response. A major problem with CD recordings is that they sometimes lack warmth, and are a little sharp (as they are too perfect). One bit tracking tries to follow the movement of the audio signal. So your CD still uses 16-bit coding.

Question: Someone told me that I couldn't use the GIF file format for my program. Are they correct?
Well it's a difficult one. In 1987 CompuServe released the GIF (Graphics Interchange Format) format as a free and open specification in 1987. It quickly became a standard way to present graphics on the WWW. Unfortunately many developers started to write software supporting GIF without even acknowledging the existence of CompuServe. Along with this GIF used a compression technique called LZW (Lempel-Ziv-Welch), which Unisys holds a patent on.
The GIF format became so successful that by at the end of December 1994, CompuServe Inc. and Unisys Corporation announced that developers would have to pay a license fee in order to continue to use technology patented by Unisys. This, though, only applied to certain categories of software supporting the GIF format. These first statements caused immediate reactions and some confusion. With all these legal discussions, it is likely that GIF will be replaced, in the future, by other formats which do not have any patent or licensing problem, especially the PNG format. The great strength of GIF over JPEG is that it supports transparent colors (which will show through the color of the background), where JPEG does not. PNG also supports this.
After a great deal of anger (including an article in Time), and with statements like:
"The announcement by CompuServe and Unisys that users of the GIF image format must register by January 10 and pay a royalty or face lawsuits for their past usage, is the online communications community's equivalent of the sneak attack at Pearl Harbor."
In the end it has been ruled the GIF file format cannot be patented, but the usage of the LZW algorithm is patented (by Unisys). So as long as you do not breach the patent for this, you are not breaching any patents. If you are you must pay a royalty for its usage.
Question: Why don't we just use the ZIP format for everything? It seems to be a good compression technique.
Well, the ZIP format uses entropy compression, which does not take into account the type of data being compressed. With audio and video compression there lots of little tricks that he can do to enhance the compression, such as:
The human eye is more sensitive to changes to luminance than it is to changes in color. Thus we can reduce the amount of information on changes of color.
The human ear tends to perceive only the music that is playing in the foreground, thus any background sounds can be suppressed.
In video compression, there tends to be very few changes from frame to frame, thus it is better to just store the difference between two frames. This can also go for audio, where the audio samples will only vary within a certain range, thus only the difference between the samples needs to be stored.
In audio and video, repetitive sequences can be located, and then reference can be made to them from previous instances, giving information on how they have changed.
Audio and video information tends to be stored in a time-based format, but if is it converted into the frequency or the spatial domain, it can considerable enhance compression, as these components tend to vary less over time. Have a look at a graphical equalizer on a hi-fi and you'll see that the frequency component in audio tends not to vary too much.
Compressed soccer results
 

Manchester Utd 2
     #16#11City4
Dundee Utd 2
     #12#73

Compression trivia - ever been stuck on a compression question in a trivia challenge?

Adaptive differential PCM codes the difference between one sample and the next. For small changes it can reduce the transmitted to one bit per sample (as the bit identifies if the sample is larger or smaller than the previous one).

What's the best letter to get in Scrabble?

Not all letters have the same probability of occurance, and this, of cours, can be used to compress data.
The league table of occurances with the points value in Scabble:
E [13.05%] - 1pt
T [9.02%] - 1 pt
O [8.21%] - 1 pt
A [7.81%] - 1 pt
N [7.28%] - 1 pt

cont --->

Ah. You say that 24-bit bit color uses 8 bits for Red, Green and Blue. So what does 32-bit color use? Is it 10 and two-thirds of a bit for each color, or have they discovered a new primary color?

No. They haven't found a new primary color. The extra 8 bits defines an alpha channel, thus every single bit can have its color set in RGB and also a transparency value from 0 to 255 (0 to 100%). Transparency allows the background color to show thorough, depending on the setting of the transparency. Graphics files, such as PNG support this, along with layers which defines the objects that are above other objects.
 
© bill@napier
 
Question: I can't understand it. I've just bought a brand-new, state-of-the-art 56kbps modem, and all I ever get is a maximum transfer speed of 4.19KB/s. Where am I going wrong, do I need a new ISP?
No. Your ISP is providing an excellent service, as 56kbps is split between send-ing and receiving. As users who access the Internet from modems typically need to receive more data than they send, the bandwidth for receiving is greater than the bandwidth for sending. You can thus receive at a faster rate than you can send. The maximum receiving rate is 33kbps, which relates to a maximum transfer rate of 4.125KB/s (there are 8 bits in a byte). If you need a higher-rate you should try ISDN which gives a total transfer rate of 128kbps (16KB/s). Otherwise consider ADSL (Asymmetric Digital Subscriber Line), which gives up to 9Mbps receiving and 1.1Mbps sending, over standard telephone lines.
Question: Everyone seems to be talking about MP-3, but what's so good about it?
MP-3 audio is set to revolutionize the way that music is distributed and licensed. A typical audio track is sampled at 44100 times per second, for two channels at 16 bits per sample. Thus the data rate is 1.411Mbps (176400B/s), giving a total of 52920000B (50.47MB) for a five-minute song. As the storage of a CD is around 650MB, it is possible to get 64 minutes from the CD.
Obviously it would take too long, with present bandwidths to download a five-minute audio file from the Internet in its raw form (over 3 hours with a 56kbps modem). If the audio file was compressed with MP-3, it can be reduced to one-tenth of its original size, without losing much of its original content.
So, it is now possible, with MP-3, to get over 10 hours of hi-fi quality music on a CD. But the big change is likely to occur with songs being sampled, and downloaded over the Internet. Users would then pay for the license to play the music, and not for purchasing the CD.
Question: Why, with video and images, do you convert from RGB into something else?
Video cameras have sensors for Red, Green and Blue (the primary colors for video information). In TV, before color TV, these colors where converted into luminance (Y). When color TV arrived they had to hide the extra color informa-tion and then send it as U and V (Redness and Blueness). Thus for TV, RGB is converted into YUV. With im-ages, the human eye is very sensitive to changes in brightness in any object, and not so sensitive to color changes. Thus color changes can be compressed more than the luminance. This is why RGB is converted in YCbCr. For example, 4:2:2 uses twice as many samples for lumi-nance than redness and blueness, and 4:1:1 uses four times as many samples.
Question: Why does MPEG have to send/store the complete picture every few frames? Would it not be possible to send/store one complete frame, and then just send/store the changes from frame to frame?
MPEG splits images up into blocks. As part of the compression process, MPEG splits each frame into a series of blocks. These blocks are then transformed. To increase compression, MPEG sends the complete picture every so often, and then just sends updates in the differences between the frames. Thus if your re-ception is not very good then you may fail to get the complete update of the picture, and only receive parts for the update. Also MPEG tries to track moving objects, it will then group the moving object, and transmit how the object moves. Sometimes this has not been encoded very well, and the object seems to move incorrectly across the screen. Normally this is because there are not enough updates to the complete frame.
This would work fine, and would give excellent compression, but the user would not be able to move quickly through the MPEG film, as the decoder would have to read the initial frame, and then all the updates to determine how the frames changed. Also if there were corrupt data, it would propagate through the whole film. Thus there is a compromise between the number of intermediate frames between each complete frame, and the number of main frames.
Compression trivia - again
 

The base rate for telephone quality speech is 64kbps, which corresponds to 8 bits at 4,000 samples per second. This is the base rate for ISDN. Any compression scheme must try and code to a multiple of this, for it to be transmitted over the phone line.
Compression trivia - again

The base rate for telephone quality speech is 64kbps, which corresponds to 8 bits at 4,000 samples per second. This is the base rate for ISDN. Any compression scheme must try and code to a multiple of this, for it to be transmitted over the phone line.
What's the best letter to get in Scrabble?

cont ...

I [6.77%] - 1 pt
R [6.64%] - 1 pt
S [6.46%] - 1 pt
H [5.85%] - 4 pts
D [4.11%] - 2 pts
L [3.60%] - 1 pt
C [2.93%] - 3 pts
F [2.88%] - 4 pts
U [2.77%] - 1 pt
M [2.62%] - 3 pts
P [2.15%] - 3 pts
Y [1.51%] - 4 pts
W [1.49%] - 4 pts
G [1.39%] - 2 pts
B [1.28%] - 3 pts
V [1.00%] - 4 pts
K [0.42%] - 5 pts
X [0.30%] - 8 pts
J [0.23%] - 8 pts
Q [0.14%] - 10 pts
Z [0.09%] - 10 pts
cont --->

 
Question: When there is JPEG, why do we still use GIF files for graphics in WWW page?
Well, the only two standard graphics files for WWW pages are JPEG and GIF. JPEG gives excellent compression, especially for photographs. It also gives 16.7 million colors, and has been designed to compress objects which have a good deal of change within the graphic. GIF files are limited in that they can only display 256 colors at a time (from a range of 16.7 colors). But GIF is very good a compressing graphics which do not have a great amount of change, especially small graphics, such as little images. The other great advantage is that GIF files support transparency, where parts of the images can be made transpar-ent, so that the background color will be show wherever there is transparency.
Question: So it GIF the future?
 
No. It's unlikely. GIF, and the compression method that it uses (LZW) are proprietary. Uni-sys hold the patent on LZW, and many software and equipment makers pay a small royalty to use it; CompuServe developed GIF and widely pro-moted it. It has allowed free use in the past but may not do this in the future.
The future is likely to be the PNG (Portable Network Graphics), which are supported by most new WWW browsers. PNG is a good format as it has strong compression for images, as there is no loss of graphic image data when an image is uncompressed. PNG also supports variable transparency of (alpha channels) and control of image brightness on different computers (gamma correction). It can be used for both small images and complex ones, such as photographs.
Question: Why when I watch digital TV, or a DVD movie, does the screen sometimes display large rectangular blocks, or objects which seem to move incorrectly across the screen?
MPEG splits images up into blocks. As part of the compression process, MPEG splits each frame into a series of blocks. These blocks are then transformed. To increase compression, MPEG sends the complete picture every so often, and then just sends updates in the differences between the frames. Thus if your re-ception is not very good then you may fail to get the complete update of the picture, and only receive parts for the update. Also MPEG tries to track moving objects, it will then group the moving object, and transmit how the object moves. Sometimes this has not been encoded very well, and the object seems to move incorrectly across the screen. Normally this is because there are not enough updates to the complete frame.
Question: All music seems to be becoming digital, but what's the great advantage when you loose something in the conversion?
Yes. Something is lost in the conversion (the quan-tization error), but this stays constant, whereas the analogue value is likely to change. The benefits of converting to digital audio outweigh the drawbacks, such as:
The quality of the digital audio system only depends on the conversion process, whereas the quality of an analogue audio system depends on the component parts of the system.
Digital components tend to be easier and cheaper to produce than high-specification analogue components.
Copying digital information is relatively easy and does not lead to a degradation of the signal.
Digital storage tends to use less physical space than equivalent analogue forms.
It is easier to transmit digital data.
Information can be added to digital data so that errors can be corrected.
Improved signal-to-noise ratios and dynamic ranges are possible with a digital audio system.
Pet of the week prize
Ever wondered what you cat looks like when compressed?
Cookie the cat with 32 colors
Cookie with 8 colors
Cookie with 4 colors

[Link]

Looking for a project in compression?

Are you interested in trying to get every last redundant bit out of your data? Well why not try an Honours project in compression. If you're interested, sign up here:
[Sign up]
Note: You must be at least 18 years old.

What's the best letter to get in Scrabble?

cont...
Thus I would say that 'h' is the best letter to get, as it has a relatively high probability, and has a high score.
And the worst ... 'u'

Looking for a project in compression?

Are you interested in trying to get every last redundant bit out of your data? Well why not try an Honours project in compression. If you're interested, sign up here:
[Sign up]
Note: You must be at least 18 years old.

How do I know which is the best for audio or video delivery? Real Audio or Windows Media Player?

Well, you could have a look at some of the tests that were made by an independent lab:
[Audio]
[Video]
For audio it can be seen that MP-3 and Real Audio easily win against the Microsoft Media format.