Images

Vector Graphics

These graphics are defined from combinations of mathematical shapes and curves. This allows them to be scaled up and down without loss of quality. Vector graphics are good for simple graphics & logos, but inefficient for photo-style graphics. Can be compressed using a lossless LZ-like compression.

File Formats

• Scalable Vector Graphics (.svg) a modern web standard as an application of XML (like HTML)
• Adobe Illustrator (.ai)
• Postscript (.ps) - Adobe format, originally used for communicating to printers/plotters and is the ancestor of the Portable Document Format (.pdf)
• Modern fonts also use vector graphics

Bitmap Images - Uncompressed (.bmp)

Bitmapped images are stored as a two-dimensional grid of pixels. "True color" images use 24 bits per pixel, 8 bits per RGB colour channel. This means that there is a simple calculation for the file size of an uncompressed image - it's basically just the volume of a box...

size in bits = (# of pixels) * (bits per pixel) = height * width * color depth

For example a 400*200 true color image:

size in kB = (400 * 200 px) * 24 bits/px * 1B/8bits * 1kB/1024B = 400*200*3/1024 kB = 234.4 kB

.bmp and .raw files are uncompressed bitmap images.

Bitmap Compression

Lossless compression: Bitmap graphics can be compressed using a LZ-like compression, but it does not compress things like photos very well. Portable Network Graphics (.png) files are a lossless format.

Lossy compression:

• Reduce resolution - simple way to reduce image quality, but good quality to compression ratio
• Reduce color space - GIFs (Graphics Interchange Format) work by using 8 bit color map instead of 24bit RGB for each pixel. This is lossy for most images. The GIF is then compressed using LZW to get further lossless compression.
• JPEG (Joint Picture Expert Group) type compression - works by hiding information that the eye does not see very well. This is based on a colour space transformation and reduction, followed by a Discrete Cosine Transform and quantization of the resultant frequency space, finally a lossless Huffman-type encoding. JPEGs get a compression ratio of about 15:1, depending on image content and compression settings.
  • See the Computerphile video for a more in depth description: JPG1 and JPG2

Audio

Midi (Musical Instrument Digital Interface)

Originally designed to be a way to connect digital instruments, but can also be used to store the sequences played/produced by a digital instrument. Midi is analogous to vector graphics and can be compressed using lossless compression.

Size of raw audio files (.wav, .aiff files)

• The sample rate is how many samples taken per second. Often measured in Hertz (Hz = 1/sec). CD quality sound uses 44.1kHz.
• The sample depth (or resolution or bit depth) is how many bits used for each sample (4 in the above image, 16 or 24 bits in CDs & DVDs)
• Bit Rate = (Sample Rate) * (Sample Depth) is the number of bits required per second.
• File size in bits = bit rate * length of recording

For example, 2 minutes of music sampled at 16000Hz with sample depth 8 bits.

Size = (120) sec * (16000 samples/sec) * (8 bits/sample) = 120*16000*8 bits = 120*16000 B = 1875 kB

Lossless Compression

FLAC (free lossless audio codec) and other lossless compression formats use things like Run-Length-Encoding, Linear Prediction, LZ Compression etc. These achieve about a 2:1 compression ratio on music.

Lossy Compression

Codecs such as mp3 (MPEG Audio Layer III), aac (advanced audio codec), wma (windows media audio), and ogg use psychoacoustics & perceptual music shaping to reduce the quality of sounds that human listeners will not perceive. If two sounds play at the same time, often the softer one can be removed. These codecs get about a 10:1 compression ratio.