Pages: « 1 [2]   Bottom of Page
 Author Topic: LAB for black and white  (Read 9466 times)
benInMA
Full Member

Offline

Posts: 186

 « Reply #20 on: February 07, 2006, 03:30:04 PM » Reply

Anyone have this book?

Photoshop LAB Color : The Canyon Conundrum and Other Adventures in the Most Powerful Colorspace (Paperback)
by Dan Margulis

Found a reference to it and it looks very interesting.

Ben
 Logged
Tim Gray
Sr. Member

Offline

Posts: 2002

 « Reply #21 on: February 07, 2006, 06:04:59 PM » Reply

It's that book that made me make the statement in my earlier post regarding interger vs floating point.  I don't recall him ever discussing the difference fp might make over integer - and his technical addendum to chapter 6 talks extensively about the quantization error - I assume if LAB used fp he would have made that point.

My google search turned up several after market products that added fp functionality in one form or another - nothing suggested that LAB used anything else than integer.
 Logged
Mark D Segal
Contributor
Sr. Member

Offline

Posts: 6752

 « Reply #22 on: February 07, 2006, 08:22:11 PM » Reply

I have the book too and I've read a good part of it. I don't think you'll find Dan Margulis agonizing over floating point math. He's far too practically-oriented for that. What I see in the book are arguments based on the logic of the processes and the comparative results you see in prints.
 Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
jani
Sr. Member

Offline

Posts: 1603

 « Reply #23 on: February 14, 2006, 08:18:31 AM » Reply

Quote
the calculations are all done internally as floating point with no loss,
This is a fallacy; floating point (or fixed point, for that matter) calculations are not lossless*. There is no such thing as infinite precision in computing.

Integer calculations, on the other hand, are lossless as long as you don't go out of bounds.

* Not all floating point or fixed point calculations are lossy, though.
 Logged

Jan
 Pages: « 1 [2]   Top of Page