Re: `Lossless' retroactive embedding of color profile (avoid JPEG recompression); basic color profile theory
Re: `Lossless' retroactive embedding of color profile (avoid JPEG recompression); basic color profile theory
- Subject: Re: `Lossless' retroactive embedding of color profile (avoid JPEG recompression); basic color profile theory
- From: Graeme Gill <email@hidden>
- Date: Wed, 02 Aug 2006 17:21:44 +1000
john castronovo wrote:
----- Original Message ----- From: "Graeme Gill"
Certainly if you continue to open and close (decompress
and compress) the same JPEG file with the same quantisation
table (usually the same program quality setting), then there
is no additional loss for each round.
To be more precise, there will be some slight changes
between successive decompress and recompress after the first,
due to numeric rounding errors in the RGB to YCrCb conversion
and DCT, that nudge a pixel value from one quantized "bin"
to the next, but these changes are small, and tend to decrease
with each round. Most of the quality loss is in the initial
compression.
Suppose the new file is saved with a better quality setting than the
first. I wouldn't expect it to get better of course, but could the
change in settings actually do some damage?
My guess is it might well change the pixel values a bit, if the
quantizing "bins" of the higher quality setting are not a strict
super-set of the lower quality settings. ie. if you're quantizing a
particular DCT coefficient into 4 "bins", and you bump up the
quality to 8 bins, nothing will change with most decoders. If you
were to change it to 5 "bins", then the change in quantized bin value,
interacting with the DCT conversion and YCbCr conversion, will increase
the number of pixel values that change (although technically the
DCT co-efficient values will not have any worse error compared to
the original uncompressed image). I think the nature of most
of the encoders quantisation tables is such that you're unlikely
to be able to choose a quality level that has quantizing values
that are a strict super-set.
Graeme Gill.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden