1. Lesson Objectives Aims You should know about: 1.3.1:
(a) Lossy vs Lossless compression.
(b) Run length encoding and dictionary coding for lossless compression.

2. Compression Means: Because: Reducing file size
Used to be down to storage limitations
Now is down to a need to make network transfers quicker

3. Compression
The level of compression is measured as a ratio between 0 and 1
1 = original file size
0 = The best compression tool ever – the delete key.

4. Compression software and algorithms
7ZIP RAR ZIP TAR TAR.GZ …

5. Example

6. How compression is achieved
Virtually all data contains “unnecessary” bits
By removing redundant information, file size (total number of bits) can be reduced
We’re basically chucking bits in the bin, however…

7. Where can data to lose come from?
Consider a video, 30fps
Does the background change?
What actually moves in that second of video?
It may actually be possible to throw away 90% of the data because it’s duplicated!

8. Lossy
Lossy compression will reduce file size but also some data will be lost
The original file/quality cannot be restored
The file size/quality is a trade off of acceptability
JPG, MP3

9. An example - Bitmap

10. File Format Differences
BMP – 644Kb
PNG – 9.05Kb (lossless!)
JPG – 30.6KB
GIF – 7.71Kb

11. Side By Side
BMP
JPG

12. Lossy Compression Why lose some data? Why not lose data?
Can be acceptable – sounds you can’t hear, un-noticable colour loss etc
Could provide much smaller file size
Why not lose data?
When loss would be unacceptable – i.e. text documents or original sound recordings

13. Lossless
Data is compressed but the file can be restored to its original quality
Methods:
Run length encoding
Dictionary coding

14. Run length encoding
There is a lot of repetition in data
Such as sequences of pixels in an image that are the same colour
Or text which has repeated letters/words.

15. Run length encoding simply counts them up and replaces them with:
A flag symbol
The character/pixel
The number of times it’s repeated
No data is lost, but file size is reduced
No point if the encoded data is the same length as the actual data!
No good in situations where there is little repetition (think of a checker board design)

16. Dictionary Encoding
Think of a 5000 word essay.
Just think of that. Mmm. Essays.
You didn’t use 5000 unique words did you?
This means there is clearly repetition

17. A dictionary of unique elements is built from the document
Then a series of pointers are created to the correct entry in the dictionary
Every time a word is repeated, data is saved.

18. Huffman Encoding
Look at the frequency of letter appearance in the English language:

19. In ASCII or UTF-8 a minimum of 8 bits is used to encode each character, yet this creates a lot of redundant data
Huffman encoding rearranges this and assigns codes to the most common letters
E = 0

20. As a result, these characters have the shortest possible codes – 1 bit!
This drastically reduces data use/size

21. Summary
Lossy compression reduces the size of a file but results in data being lost.
Lossless compression reduces the file size without losing any data.
Run-length encoding replaces a sequence of repeated characters with a flag character, followed by the character itself and the number of times it is repeated.
In dictionary coding a dictionary of commonly occurring sequences of characters is created. In the text these sequences are replaced by pointers to the relevant place in the dictionary. The references are shorter than the words they replace, so the file is reduced in size.

22. Review/Success Criteria