From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: From: "Steve Simon" Date: Fri, 27 Oct 2006 09:46:43 +0100 To: 9fans@cse.psu.edu Subject: Re: [9fans] Proposed: Image Conversion Server In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Topicbox-Message-UUID: d768836c-ead1-11e9-9d60-3106f5b1d025 >> surely you'd want to undo the premultiply of the alpha Yes I believe you do, the alpha blends the image with others it is to be composited with, for image processing you want to undo this and then redo it after processing - say you want to threshold the image to into just two levels of grey, now your blending is undone unless you firs= t remove the blending, do the threshold and then replace it. > http://alvyray.com/Awards/AwardsAcademy96.htm > unfortunately i'm unable to currently locate the original paper referen= ced at the bottom > of that page. There is a link to it at the bottom of this page =E2=98=BA http://plan9.bell-labs.com/wiki/plan9/Papers/index.html > maybe there's something i'm missing, but what is the benefit to using d= ifferent colorspaces? In theroy you can do any manipulation in any colourspace but if you want = to say, correct spatial luma variations in an image and adjust its saturatio= n then having it in CMYK is much more efficent than RGB, you do get the ove= rhead of conversion of course, and if you don't keep enough bits in the middle the= n you get colourspace quantisation. Swings vs. roundabouts as ever. -Steve