[gs-devel] How to tell when DCTEncode is OK? (re #450663)

[L. Peter Deutsch]

> Encode source data, decode the result and compare 2nd result with source
> data.
> Image is a vector of color components, so the comparison is a norm and a
> threshold.
> Not sure which norm is best though.
> Not sure if it is fast enough.

I think simply using a threshold for the difference is not sufficient.  One
of the typical bad results from inappropriate DCT encoding of is a lot of
faint "speckling" in the decoded image.  The individual differences are
small: it's the loss of monotonicity that is objectionable.

I liked Raph's suggestion and will try using it when I get around to this
project.

-- 

L. Peter Deutsch     |    Aladdin Enterprises   |  203 Santa Margarita Ave.
ghost at aladdin.com    |  http://www.aladdin.com  |  Menlo Park, CA 94025

   No government censorship of software!  Oppose the CBDTPA!  www.eff.org