Re: proposal: background-image-alpha

----- Original Message ----- 
From: "David Woolley" <>
To: <>
Sent: Monday, May 08, 2006 11:25 PM
Subject: Re: proposal: background-image-alpha

| > This RGB to grayscale translation is a subject of holy wars on the Net, like 
| > it
| > luminance or what... And what if this GIF contains transparent color...
| We are talking about zero chrominance, so to get a grayscale image with
| the same gamma as the the full image, one simply selects any channel,
| as they are all the same.

"they are all the same". How will you know this in case of GIF or RGB JPEG
for any arbitrary image? The only consistent formats for alpha are grayscale
subformats of JPEG and PNG.

For GIFs something like luminance-to-alpha needs to be defined in some other 

| Someone else mentioned gamma issues with PNG.  I'm not familiar with this
| problem, but I am aware that one of the most common defects on the web
| is using images in GIF and JPEG which are corected for a gamma of around 1.0 
| around 1.6 (Mac) when images for the web should be sRGB (gamma 2.2) images, so
| I would say that the designers of most web pages don't understand gamma.
| (Also a lot of text anti-aliasing fails to temporarily convert to gamma 1
| before averaging.)
| Of course, alpha masks are not grayscale images, so if one uses a nominally
| grayscale image format for them, I would expect the software to actually
| assume a linear encoding for the mask.

What does "linear encoding for the mask" mean here?

| > It is better to use just PNG and JPEGs (JPEG 2000) with alpha to be short.
| Not at the moment, and in the time frames of new CSS standards, one cannot
| assume what image formats will be available, as the actual formats don't 
| only the capabilities.

At least PNG is here. And even IE can do transparency on it these days.

Andrew Fedoniouk.

Received on Tuesday, 9 May 2006 20:31:39 UTC