Quantisation
- Each DCT component divided by quantisation value and rounded to nearest integer
- Quantisation biased to preserve low frequencies and remove high frequencies
- A stronger (more lossy) quantisation table may be used to increase the compression
- Quantisation table is stored in JPEG image
- Inverse process: multiply each component by stored quantisation value
- Components stored in quantised form since smaller numbers mean less bits
[Next] [Contents]