Hey guys Iam trying to combine a lowres indirect illumination texture of 256x256 pixels with a 1024x1024 highres texture. I saw this article here : http://developer.download.nvidia.com/presentations/2010/futuregameon/LouisBavoil_ModernRealTimeRenderingTechniques.pdf
On page 20, a technique from a game is described to combine the lowres particels of the game with the fullres rendering. Well I implemented it the same way, but I don't get a realy good result. I use 4 samples which are defined like this : float2( tex.x - lowres_texel,tex.y ) , float2( tex.x + lowres_texel, tex.y) , float2( tex.x, tex.y + lowres_texel) , float2( tex.x, tex.y - lowres_texel), where tex is the texcoord of the current highres pixel and lowres_texel = 1/256.
Then I going through a for loop and combine the new depthdistance to the highresdepth with the previouse ones and save the index of the sample, which has the nearest depth of all. After that I look if the sample is in the range from 0-3( that would be the 2x2 samples, and when -1 is sampled only with the standart texcoord tex) and output the color from this sample.
How could I improve it ?? Because in the paper it works realy good