Shadowmapping

Czarek :) · 2014-12-14T19:06:05

Hi there i try to implement shadowmapping, hwever i get something like this: So far i managed to see if i get proper depth map (and yes i have good depth texture) So i think problem is with mapping vertex/fragment coords into shadowmap texture space and reading its z value (not from shadowmap) Maybe first ill explain how i am doing everything: (or at least i try to do) I set viewport to 1024x512, set the same shadowmap texture size, set perspective projection matrix to znear 10 z far 1000 000. then i do actual drawing: Clear depth and color buff Render the scene from light POV, copy depth data to the texture Clear depth and color buff Render the scene from camera POV and send shadowMVPmatrix * texturebias matrix to the shader: Matrix44<double> SMVP = ( (ACTUAL_MODEL * shadowView) * ACTUAL_PROJECTION ) * ShadowbiasTex; where ShadowbiasTex is void LoadBiasIdentity() { (*this) = Matrix44<T>( 0.5, 0.0, 0.0, 0.0, 0.0, 0.5, 0.0, 0.0, 0.0, 0.0, 0.5, 0.0, 0.5, 0.5, 0.5, 1.0); Transpose(); } Now for the shadowmapping vertex and frag shaders: in Vertex Shader: Compute VertexScreenPos = MVPmatrix * vertex; Compute VertexShadowScrPos = SMVPTmatrix * vertex; Send VertexShadowScrPos to fragment shader output vertex color = 1,1,1,1 texcoords are passed not touched in Fragment Shader VertexShadowScrPos.z = VertexShadowScrPos.z / VertexShadowScrPos.w; float DepthInLightPOV = GetTextureSampleFrom(VertexShadowScrPos.x, VertexShadowScrPos.y); add little tiny small offset to it DepthInLightPOV = DepthInLightPOV + 0.0000625; float RESULT_COLOR = DepthInLightPOV - VertexShadowScrPos.z; clamp result 0..1 and multiple it by 100000 and clamp again to 0..1 <-- depth compare if RESULT_COLOR is less than 0 then we are in shadow if not we are not in the shadow Set fragment color as RESULT_COLOR Someone can ask for the code code is bad and will liekly rip off your eyes but here it is: OpenGL rendering pass: void Draw() { //we start with activetexture 0 glClear(GL_COLOR_BUFFER_BIT); glClear(GL_DEPTH_BUFFER_BIT); if ( glIsEnabled(GL_DEPTH_TEST) == false) glEnable(GL_DEPTH_TEST); gluPerspectiveA(90.0, 2.0, 10.0, 1000.0 * 1000.0); ACTUAL_MODEL.LoadIdentity(); FPP_CAM->SetView(); //sets ACTUAL_VIEW //shadowmapping if (village != NULL) { glActiveTextureARB(GL_TEXTURE1_ARB); //texturing already enabled (was enabled during initialization) if ( glIsEnabled(GL_DEPTH_TEST) == false) glEnable(GL_DEPTH_TEST); glEnable( GL_VERTEX_PROGRAM_ARB ); VertexOnly_ID.Bind(); SetShaderMatrix(ACTUAL_MODEL, shadowView, ACTUAL_PROJECTION); village->DrawSimpleModel(); glDisable( GL_VERTEX_PROGRAM_ARB ); glReadBuffer(GL_DEPTH_BUFFER_BIT); glBindTexture(GL_TEXTURE_2D, shadowtex); glCopyTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, 0, 0, 1024, 512); //depth tex copied glActiveTextureARB(GL_TEXTURE0_ARB); glClear(GL_COLOR_BUFFER_BIT); glClear(GL_DEPTH_BUFFER_BIT); glEnable(GL_TEXTURE_2D); glEnable( GL_VERTEX_PROGRAM_ARB ); glEnable( GL_FRAGMENT_PROGRAM_ARB ); SHADOW_SHADER.Bind(); SetShaderMatrix(ACTUAL_MODEL, ACTUAL_VIEW, ACTUAL_PROJECTION); Matrix44<double> biasmvp; biasmvp.LoadBiasIdentity(); Matrix44<double> SMVP = ((ACTUAL_MODEL * shadowView) * ACTUAL_PROJECTION)*biasmvp; SendMatrixToShader(SMVP, 5); village->DrawSimpleModel(); glDisable( GL_FRAGMENT_PROGRAM_ARB ); glDisable( GL_VERTEX_PROGRAM_ARB ); glDisable(GL_TEXTURE_2D); }} now things will be messy shaders! ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- !!ARBvp1.0 PARAM MVP1 = program.local[1]; PARAM MVP2 = program.local[2]; PARAM MVP3 = program.local[3]; PARAM MVP4 = program.local[4]; PARAM SMVP1 = program.local[5]; PARAM SMVP2 = program.local[6]; PARAM SMVP3 = program.local[7]; PARAM SMVP4 = program.local[8]; #compute vertex position TEMP vertexClip; DP4 vertexClip.x, MVP1, vertex.position; DP4 vertexClip.y, MVP2, vertex.position; DP4 vertexClip.z, MVP3, vertex.position; DP4 vertexClip.w, MVP4, vertex.position; #computre vertex in light space position TEMP shadowClip; DP4 shadowClip.x, SMVP1, vertex.position; DP4 shadowClip.y, SMVP2, vertex.position; DP4 shadowClip.z, SMVP3, vertex.position; DP4 shadowClip.w, SMVP4, vertex.position; #set color as rgb 1,1,1,1 TEMP COL; MOV COL.x, 1.0; MOV COL.y, 1.0; MOV COL.z, 1.0; MOV COL.w, 1.0; MOV result.color, COL; #send vertex pos to oglpipeline MOV result.position, vertexClip; #texture coords untouched MOV result.texcoord[0], vertex.texcoord; #send vertex position in light view to fragment MOV result.texcoord[2].x, shadowClip.x; MOV result.texcoord[2].y, shadowClip.y; MOV result.texcoord[3].x, shadowClip.z; MOV result.texcoord[3].y, shadowClip.w; END ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Fragment shader: !!ARBfp1.0 TEMP ShadowDEPTH; #get the data we send to fragment shader TEMP ShadowClip; MOV ShadowClip.x, fragment.texcoord[2].x; MOV ShadowClip.y, fragment.texcoord[2].y; MOV ShadowClip.z, fragment.texcoord[3].x; MOV ShadowClip.w, fragment.texcoord[3].y; TEMP W_DIV; #Make W_DIV.x = 1.0 / ShadowClip.w; RCP W_DIV.x, ShadowClip.w; # shadowclip.z = shadowclip.z / shadowclip.w; MUL ShadowClip.z, ShadowClip.z, W_DIV.x; #make a 2d point for texture sampling just to be sure TEMP ShadowCoord; MOV ShadowCoord.x, ShadowClip.x; MOV ShadowCoord.y, ShadowClip.y; MOV ShadowCoord.z, 0.0; MOV ShadowCoord.w, 1.0; #take texture sample out of ShadowCoord x and y TEX ShadowDEPTH, ShadowCoord, texture[1], 2D; #add offset ADD ShadowDEPTH.x, ShadowDEPTH.x, 0.0000625; MAX ShadowDEPTH.x, ShadowDEPTH.x, 0.0; MIN ShadowDEPTH.x, ShadowDEPTH.x, 1.0; TEMP dRES; #compute if we are in shadow by subtracting Depth_FROM_TEX - Vertex_in_shadow.z SUB dRES.x, ShadowDEPTH.x, ShadowClip.z; #if its smaller than 0 clamp it to 0 MAX dRES.x, dRES.x, 0.0; #if its bigger than 0 multiple it y a big number and then clamp to 1 to get 1 MUL dRES.x, dRES.x, 100000.0; MIN dRES.x, 1.0, dRES.x; #output the color MOV result.color.r, dRES.x; MOV result.color.g, dRES.x; MOV result.color.b, dRES.x; MOV result.color.a, 1.0; END ================================================================================================================== Again results LIGHT POV camera pov another shot from cam pov

Graphics and GPU Programming Programming OpenGL

Started by _WeirdCat_ December 07, 2014 11:09 PM

10 comments, last by _WeirdCat_ 9 years, 4 months ago

jeskeca

676

December 14, 2014 01:04 PM

Peter panning happens with a *thin* wall because there is not enough z-precision for a thin wall to cast a shadow on the ground immediately behind it. (You can try a 32bit shadowmap, but that still might not be enough precision.

A *fat* wall does not have this problem, as the wall is thicker than the z-precison.

The article suggests the fix is easy, tell your artists not to make *thin* geometry.

_WeirdCat_

1,512

Author

December 14, 2014 07:06 PM

i still dont get the idea of not making thin gemoetry athoug i understand waht you mean however my heightmap data is 'thin geometry ' ps i dont have an artist :P

SO i think the best option would be correcting z_near and z_far for close objects and making another rendering pass with differnt near./far for object that are further

Master & Mentor
https://sites.google.com/site/customprog/

Shadowmapping

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Shadowmapping

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines