Many years ago I reported a problem in OI16 with images on PCs set up with 256 colours. The problem might not have appeared on all systems with 256 colours (or else I would have got more complaints from my clients). Prompted by the recent posting by Alberto Arreola Pérez (which might be about a quite diffent problem, being written in a Spanish quite above my very basic knowledge of that language (cerveza, playa, Almodovar)) I did some experiments.

On a Win NT PC with 256 colours OI16 (OI 3.7.1) displays the images as well as can be expected. But on a different PC with Win2K I can clearly see the problem. And it turns out the problem persists in OI32 (OI 4.1.2)!

Here are my results on Win2K. An additional symptom for the problem is that the images show as black rectangles at design time.

Could the problem lie somewhere else than in OI16?

It'll be interesting to see if others can reproduce the problem. Here and here are my two test files.

OI16 (3.7.1):

Win2K set up with 24-bit colour (32 bits actually):

Win2K set up with 256 colours:

OI32 (4.1.2):

Win2K set up with 24-bit colour (32 bits actually):

Win2K set up with 256 colours:

- Oystein -

The thing about the palette issue (256 colors) is that Windows has to figure out a generic palette that works for all applications that are visible. How this is done depends upon the version of Windows. After a generic palette is created, the visible images then are rendered with this palette. That means mapping the image to the palette using some sort of color fitting algorithm.

A normal (expected) result is for the colors to be slightly off and either mapped to the closest matching color or else dithered so that the average of several pixels matches the average true colors for those same pixels. Dithering can be done either by the application or by Windows.

Starting with Windows 2000 you also have color management.

Victor,

I'll do some more experiments in light of what you say, with my own images, and perhaps one of Alberto's. But I must confess I find it hard to believe that the stark yellow/red/black version of my image shown at the bottom of my posting is the result of palette problems and nothing else. Some of the other images do show signs of palette related distortions, I agree.

- Oystein -

Victor,

Thanks.

I did some experiments in Win2K set up with 256 colours. In my experiments I ran several OI32 windows, one at a time, each containing one image. Mostly the colours came out bad.

For each window I maximized, minimized and restored the window, to give Windows a chance to recompute its palette. This made no difference. (Please tell me if there was something else I should have done to help the colours come out right.)

I found the images' colours were wrong in different ways.

A couple of images with a web palette (e.g, the netcolsm image from one of your web pages) came out with the colours seemingly snapping to the nearest colour in the old Windows 16 colour palette. It seems to me something (Windows? OI32???) should have done a better job than collapsing to 16 colours. Can you think of an explanation why this happened?

An image with an "exact" palette (in PhotoShop's terminology), i.e, a palette assumingly consisting of the "best" colours from a 24-bit original, came out mostly good, but some little used (???) colours were quite off. (I.e, most pixels seemed to be right, but the ones that were off ruined the whole image.)

An image with an "Windows System" palette came out fine.

But I also had the aforementioned "exact" and "Windows System" images tripping out on acid, like a totally unrelated palette was used. Even if I minimized and restored them, as mentioned earlier. PhotoShop was in the background, though. I don't know if that is relevant.

I look forward to your reply.

- Oystein -

The bottom one looks like it may by trying to use a 4 bit palette.

Photoshop being in the background might have been relevant. I don't know. What I would do is load OI as the only app. Display the problem image. If you have the problem at this point, the problem is probably not a result of Windows paletting calculation issues.

One thing that you might want to do is to use an application that can display the palette and watch what happens to it directly, as you switch between applications. With the palette displayed in palette order, you can watch how it is remapped as it's being remapped.

Victor,

Thanks. I might do more experiments tomorrow. I'm one a different computer here at home, which can't be set lower than 16-bit colours. Bliss.

Tell me - do you think I would have to reboot Windows or restart OI after changing the colour setting, to make the experiments valid? I haven't done so far.

- Oystein -

I don't think that should be necessary. Also, I can't believe you can't set below 16 bit. Try using Multires (available at http://www.entechtaiwan.com), or Powerstrip.

Victor,

I'll put it off until tomorrow. Thanks.

¹⁾)

- Oystein -

Victor,

I had to put the colour problems away for a while. But now I tried the netcolsm.bmp from your web pages again and still got only 16 colours.

Here is what I do:

- Run OI 4.1.2 on Win2K with 256 colours…

- …with no other programs running.

- Test run an OI form containing the netcolsm.bmp image (clipped)…

- …or rather I use a larger, resized version of that image, but that shouldn't matter.

Here is how it looks:

Here is how it looks when I run Windows in full colour:

Can anything be concluded from these facts?

- Oystein -

Victor,

Have you got any comments to my previous posting on this subject?

It's not an important matter for me, though. In my apps I have images two kinds of places - (1) buttons and the like, and (2) bitmap controls where clients' images are shown. For (1) I can avoid the problem by using the right formats and palettes (although sometimes I need to cram as much detail as possible into small areas to make the symbols understandable for the client). (2) is a non-problem since clients that have images have their PCs set up with more than 256 colours.

- Oystein -

Isn't it interesting how if your name appears on a page you see it right away? Your change of the subject definitely got my attention!

Anyway, judging from others' contributions to this thread, it appears this may be an OI artifact and not a Windows problem like I originally suggested. Not knowing how OI processing images, I choose not to speculate further.

Those 16 colors, by the way, look like the standard DOS/Windows 16 colors.

Victor,

Hey - you changed the subject back!

Thanks. It's a matter for Revelation then. But perhaps not worth pursuing.

Wonder how it worked out with Alberto Arreola Pérez.

- Oystein -

View this thread on the forum...