Understanding Wavelet Compression

From Redhax

[edit] What is Image Compression

Run Length Encoding (RLE) is one example of a method of image compression that works well for simple images with many pixels of the exact same color.

Compression is the process of reducing the amount of data required to represent a sample of data.

Compression falls under one of two categories. Lossless and Lossy.

In a lossless compression scheme the data which is compressed will later be perfectly reconstructed. In a lossy compression scheme data that is determined to be "less important" or unmissed is discarded. An example of Lossy compression would be taking the sentence: "I took a bus downtown this afternoon to meet up with Jane and we had a nice lunch"

and compressing it to: "I ate lunch with Jane downtown today".

The compression scheme employed reduced the sentence by almost half but maintained pretty much all of the important information: who, what, when and where but some of the nuance was lost in the process. Similarly with lossy image compression the goal is to maintain as much nuance as possible while not losing the important details.

[edit] Discrete Cosine Transform CompressionThe engine behind HD, DV, Jpeg and many others

[edit] Wavelet Compression

[edit] How Wavelet Compression Works

[edit] Comparison to other methods

[edit] When should you choose not to compress?

Sometimes, the process of de-compressing the frames takes more time than moving the uncompressed data around, and this is important when speed (of exporting frames e.g.) has a higher priority than disk-space.

Lossy compression should only be used, IMHO, when no Visual FX (or color correction) will be applied, or as the last format in the workflow chain.