[SOLVED] Extracting pixel's worldspace position

[devsh]

Eureka.... I found extracting worldspace pos has been done before in SSAO in Xeffects... all i need to do is copy and adapt.

[devsh]

I think I made it

Code: Select all

// simple fragment shader
uniform sampler2D DM;
uniform float FarValue;
uniform mat4 reverseScreenVP;
uniform float xDist;
uniform vec4 point;

void main()
{
	vec2 texcoord = vec2(gl_TexCoord[0]);
	vec4 texDepth = texture2D(DM, texcoord);
	float depth = (texDepth.r + texDepth.g / 256.0)*FarValue;
    texcoord -= vec2(0.5,0.5);
    texcoord *= 2.0;
    texcoord.y = - texcoord.y;
	vec4 ScreenPos = vec4(texcoord.x*depth, texcoord.y*depth, depth, 1.0);
    ScreenPos = reverseScreenVP*ScreenPos;
    if (ScreenPos.y > xDist) gl_FragColor = vec4(1.0);
}

The white bits correspond to world space coords

[devsh]

wrong doesnt work at all....

something I need to do with matrices maybe

[devsh]

ok I think I am able to retrieve the screen position coords (the one after multiplying model space gl_Vertex with gl_ModelViewMatrix)

the shader I used for terrain (quick mod)

Code: Select all

uniform float MaxD;
uniform float s;
uniform float s2;
uniform vec4 sunDiffuse;
uniform sampler2D Tex;
uniform sampler2D Tex2;
varying vec4 pos;
varying vec3 normal;
varying vec4 Sunpos;
varying vec4 Moonpos;
varying float night;

void main() 
{
	float depth = pos.z / MaxD;
	float mulDepth = depth * 256.0 + 2.0;
	float flooredDepth = floor(mulDepth);

	float shade = max(dot(Sunpos,normal),0.0)*night;
	float moonlit = max(dot(Moonpos,normal),0.0)*(1.0-night);

	vec4 tex = texture2D(Tex, gl_TexCoord[0].xy*s) * texture2D(Tex2, gl_TexCoord[0].xy*s2);
        float diffuse = gl_FrontMaterial.diffuse*sunDiffuse;
	vec4 globalAmbient = tex*gl_LightModel.ambient * gl_FrontMaterial.ambient;

	gl_FragData[0] = vec4(vec3(pos/MaxD),1.0)/*vec4(tex.rgb*shade*diffuse+globalAmbient + tex.rgb*moonlit*gl_FrontMaterial.ambient,1.0 )*/;
	gl_FragData[1] = vec4(flooredDepth / 256.0, mulDepth - flooredDepth, 0.0, 1.0 );
}

And the shader I used in Shader Maker

Code: Select all

// simple fragment shader
uniform sampler2D DM; // DM for Depth Map
uniform sampler2D color;
uniform float FarValue;
uniform mat4 reverseScreenVP;

void main()
{
	vec2 texcoord = vec2(gl_TexCoord[0]);
	vec4 texDepth = texture2D(DM, texcoord);
	float depth = (texDepth.r + texDepth.g / 256.0)*FarValue;
    texcoord -= vec2(0.5,0.5);
    texcoord *= 2.0;
    texcoord.y = - texcoord.y;
	vec4 ScreenPos = vec4(texcoord.x*depth, texcoord.y*depth, depth, 1.0);
    //ScreenPos = reverseScreenVP*ScreenPos;
    ScreenPos /= FarValue;
    gl_FragColor = ScreenPos;
}

as you can see they do match up

[devsh]

SO I just need confirmation if ANYONE BOTHERS to reply...

did I manage to retrieve screen space position?

[devsh]

just another way I thought of

Code: Select all

// simple fragment shader

// 'time' contains seconds since the program was linked.
uniform vec3 norm0,norm1,norm2,norm3, cameraPos;
uniform sampler2D DepthMapSampler, col;
uniform float FarLink;

vec3 getScreenNormal( void ) {
	vec3 left = mix(norm1-cameraPos, norm0-cameraPos, gl_TexCoord[0].y);
	vec3 right = mix(norm3-cameraPos, norm2-cameraPos, gl_TexCoord[0].y);
	return mix(left, right, gl_TexCoord[0].x);
}

float getDepth( void )
{
	vec4 texDepth = texture2D(DepthMapSampler, gl_TexCoord[0].xy);
	float extractedDepth = (texDepth.r + texDepth.g / 256.0);
	return extractedDepth;
}

void main()
{
	
	vec3 pos = getScreenNormal()*getDepth()/*FarLink*/+cameraPos;
	if (pos.y >700.0) gl_FragColor = texture2D(col,gl_TexCoord[0].xy);
}

[devsh]

IDIOTIC and dead simple

Code: Select all

uniform vec3 norm0,norm1,norm2,norm3, cameraPos;
uniform sampler2D DepthMapSampler, col;

vec3 getScreenNormal( void ) {
	vec3 left = mix(norm1, norm0, gl_TexCoord[0].y);
	vec3 right = mix(norm3, norm2, gl_TexCoord[0].y);
	return mix(left, right, gl_TexCoord[0].x);
}

float getDepth( void )
{
	vec4 texDepth = texture2D(DepthMapSampler, gl_TexCoord[0].xy);
	float extractedDepth = (texDepth.r + texDepth.g / 256.0);
	return extractedDepth;
}

void main()
{
	
	vec3 pos = mix(cameraPos, getScreenNormal(), getDepth()/7);
	if (pos.y >2150.0) gl_FragColor = texture2D(col,gl_TexCoord[0].xy);
}

[Nadro]

Why You don't use "Edit" function? Forum isn't a trash...

[devsh]

sorry i was a bit overexcited

[BlindSide]

Heh cool that's the way I used in the end too, passing 4 direction vectors and lerping based on screen pos. Nice. If you look at the shader source for XEffects SSAO you can see it there, would have saved you some time.

Wait whats with the "vec3 pos = mix(cameraPos, getScreenNormal(), getDepth()/7); ", shouldn't you be doing "vec3 pos = cameraPos + getScreenNormal() * getDepth() * farValue;"??

[devsh]

i had a look at your source code.. but it kinda didnt work for me and my method is a bit faster. I dont have to substract camPos from each position of the far plane to make a normal. I tell you how it works

After you interpolated a coordinate in WSpace (not a normal) on the far plane...
Imagine a line between that point and camera position, and a slider going between them. I use the mix function to "move the slider" and its ideal since the depth read off the depthMap is clamped in 0.0 to 1.0;

The /7 was there because my depth map was made with a MaxDistance of 10000 when my far value was 70000

[devsh]

Just to share with you my success



I used reconstruction of pixel coords to do that in the water shader, it casts caustics on everything... No more playing "Receiver", "emitter"

[Lonesome Ducky]

Looking pretty good there devsh

[sio2]

Imagine a line between that point and camera position, and a slider going between them. I use the mix function to "move the slider" and its ideal since the depth read off the depthMap is clamped in 0.0 to 1.0;

Depth isn't linear?

[devsh]

how about it works cause i have tested retrieving positions and heights