image processing

[x265]

Could someone explain what quantization means?

[Groucho2004]

Could someone explain what quantization means?
You mean the 2.4 million hits on Google for "image processing quantization" are not enough?
Do you want fries with that?

[x265]

Quantization, involved in image processing, is a lossy compression technique achieved by compressing a range of values to a single quantum value. When the number of discrete symbols in a given stream is reduced, the stream becomes more compressible. --Wikipedia

I don't understand this.

[Groucho2004]

I don't understand this.
How about this (http://www.google.co.uk/url?sa=t&rct=j&q=&source=web&cd=3&cad=rja&ved=0CEEQFjAC&url=http%3A%2F%2Fwww.cs.princeton.edu%2Fcourses%2Farchive%2Ffall00%2Fcs426%2Flectures%2Fdither%2Fdither.pdf&ei=J4J6UenkJaON7QahvoC4CQ&usg=AFQjCNH9fZ9wmpeXES1J2lv94z-OFO9Igg) article?

[Asmodian]

Well those articles do not help me understand it better.

I think of simple quantization, in the context of MPEG4 video compression, as dividing the matrix of coefficients by a number.

This allows the arithmetic encoder to use fewer bits when compressing this matrix due to the lower precision. It takes fewer bits to save 177/4=44 vs 177 and we don't notice the weight being off by 1 (44*4=176) after converting back and then doing the inverse DCT.

The matrix is the weights for the the different frequencies in the vertical and horizontal directions from the image converted to frequency space (after the DCT). Thinking of images in frequency space is hard (for me at least). :o

The DCT Wikipedia page (http://en.wikipedia.org/wiki/Discrete_cosine_transform) is pretty good if you want to understand what this matrix is from.