Question

I am displaying a texture that I want to manipulate without out affecting the image data. I want to be able to clamp the texel values so that anything below the lower value becomes 0, anything above the upper value becomes 0, and anything between is linearly mapped from 0 to 1.

Originally, to display my image I was using glDrawPixels. And to solve the problem above I would create a color map using glPixelMap. This worked beautifully. However, for performance reasons I have begun using textures to display my image. The glPixelMap approach no longer seems to work. Well that approach may work but I was unable to get it working.

I then tried using glPixelTransfer to set scales and bias’. This seemed to have some sort of effect (not necessarily the desired) on first pass, but when the upper and lower constraints were changed no effect was visible.

I was then told that fragment shaders would work. But after a call to glGetString(GL_EXTENSIONS), I found that GL_ARB_fragment_shader was not supported. Plus, a call to glCreateShaderObjectARB cause a nullreferenceexception.

So now I am at a loss. What should I do? Please Help.

---

What ever might work I am willing to try. The vendor is Intel and the renderer is Intel 945G. I am unfortunately confined to a graphics card that is integrated on the motherboard, and only has gl 1.4.

Thanks for your response thus far.

Answer 1

Unless you have a pretty old graphics-card, it’s surprising that you don’t have fragment-shader support. I’d suggest you try double-checking using this.

Also, are you sure you want anything above the max value to be 0? Perhaps you meant 1? If you did mean 1 and not 0 then are quite long-winded ways to do what you’re asking.

The condensed answer is that you use multiple rendering-passes. First you render the image at normal intensity. Then you use subtractive blending (look up glBlendEquation) to subtract your minimum value. Then you use additive blending to multiply everything up by 1/(max-min) (which may need multiple passes).

If you really want to do this, please post back the GL_VENDOR and GL_RENDERER for your graphics-card.

Edit: Hmm. Intel 945G don’t have ARB_fragment_shader, but it does have ARB_fragment_program which will also do the trick.

Your fragment-code should look something like this (but it’s been a while since I wrote any so it’s probably bugged)

!!ARBfp1.0 ATTRIB tex = fragment.texcoord[0] PARAM cbias = program.local[0] PARAM cscale = program.local[1] OUTPUT cout = result.color  TEMP tmp TXP tmp, tex, texture[0], 2D SUB tmp, tmp, cbias MUL cout, tmp, cscale  END  

You load this into OpenGL like so:

GLuint prog; glEnable(GL_FRAGMENT_PROGRAM_ARB); glGenProgramsARB(1, &prog); glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, prog); glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, strlen(src), src); glDisable(GL_FRAGMENT_PROGRAM_ARB); 

Then, before rendering your geometry, you do this:

glEnable(GL_FRAGMENT_PROGRAM_ARB); glBindProgramARB(GL_FRAGMENT_PROGRAM_ARB, prog); colour4f cbias = cmin; colour4f cscale = 1.0f / (cmax-cmin); glProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 0, cbias.r, cbias.g, cbias.b, cbias.a); glProgramLocalParameter4fARB(GL_FRAGMENT_PROGRAM_ARB, 1, cscale.r, cscale.g, cscale.b, cscale.a);  //Draw your textured geometry  glDisable(GL_FRAGMENT_PROGRAM_ARB);