compression-techniques

File compression

MPEG-3 (MP3) – Audio compression

• MP3 uses audio compression to reduce file size

• A typical music file can be reduced by up to 90% (e.g. 80 MB to 8 MB)

• MP3 files are used on:

  • MP3 players

  • Computers

  • Smart phones

• Music can be:

  • Downloaded or streamed from the internet

  • Converted from a CD to MP3 format

• MP3 quality isn’t as high as the original CD, but it’s good enough for most people

• MP3 uses perceptual music shaping to remove sounds we don’t notice:

  • Frequencies outside human hearing range

  • Quieter sounds that are masked by louder ones

• This makes the file smaller without a big drop in quality

• MP3 is a lossy format:

  • Some original data is permanently lost

  • You can’t get the original file back

MPEG-4 (MP4) – Multimedia file format

• MP4 is similar to MP3, but stores more than just audio

• It can store:

  • Music

  • Videos

  • Photos

  • Animations

• Commonly used for streaming videos over the internet

• MP4 uses compression to keep file size small

• Maintains high quality without noticeable loss to the viewer

Images

• Bitmap images can be compressed, the file size and quality are reduced

• JPEG is the most common file format for bitmap images

• The JPEG compression algorithm creates a new file, so the original file can no longer be used

• Vector graphics can also be compressed but with limited success

• Scalable vector graphics (.svg) use an XML text file which means that allows them to be compressed.

Run-length encoding (RLE)

What is run-length encoding?

• Run-length encoding (RLE) is a form of data compression that condenses identical elements into a single value with a count

Text data

• For text data containing the string “AAAABBBCCDAA“, the plain RLE encoding would be “4A3B2C1D2A“

• The string has:

  • four ‘A’s (4A)

  • three ‘B’s (3B)

  • two ‘C’s (2C)

  • one ‘D’ (1D)

  • two ‘A’s (2A)

• To represent this in binary, the count is stored in a fixed size binary format (e.g. 7 or 8 bits)

• The character is stored using its ASCII value (7 bits)

• The binary RLE representation of 4A would be 0000100 1000001

  • 0000100 – binary for the count (4)

  • 1000001 – binary for ‘A’ (65)

Images

• Run-length encoding (RLE) uses frequency/data pairs to compress bitmap image data

• For example, the following bitmap image with a bit depth of 1 bit would have the following binary bit pattern

Bitmap	Bit pattern

• Using RLE we group pixel colours and can create frequency/data pairs as follows

  • 30, 11, 20, 11, 20, 11, 50, 11, 30, 11, 10, 11, 60, 11……

• 3 x 0 = 3 x white, 1 x 1 = 1 x black

• Data pairs can carry over on to the next line, e.g. end of first line and start of second line is 5 x 0 (5 x white)

• The original file would need to use 195 bits of storage space

• The RLE compressed file would need use 38 bits of storage space, a saving of 157 bits (80%)

• For images using more than two colours, the RGB value for each colour is stored along with the count

• For example, an image using the following four colours:

• The data stored could look like:

  • 10 0 0 0 5 255 0 0 3 0 255 0

• Ten black pixels ((10) 0 0 0) – count + RGB

• Five red pixels ((5) 255 0 0) – count + RGB

• Three green pixels ((3) 0 255 0) – count + RGB