[EDIT :] I just realize the the wrong forum to post it (it would be better in Graphics, sorry) Hi, let summary, I have a shadow map and a depth value corresponding to the projection of a 3D point in the light space. The depth value is the actual value, but the value picked in the shadow is dependent of the shadow map resolution. The error between the two value is big each time there is a depth step (you can see the pictures if it's not clear, the links are below). Moreover, the error value is more or less equal to the depth step value. I don't have a real solution, I try to quantify the error to use a epsilon, but the major problem is the foreshortening, I am thinking about validating the point if the depth is in the range defines by the [min,max] value of the shadowMap's corresponding pixel's neighbourhood (But that implies 9 texture accesses :( ).(I use a fragment shader) So anyone with any comments or links is welcome. Thanks, -uto- 3D point depth variation in the light space [link]http://members.gamedev.net/uto/ShadowMap/PointDepthGradient.jpg[/link] Shadow depth variation [link]http://members.gamedev.net/uto/ShadowMap/shadowDepthGradient.jpg[/link] Difference between the two [link]http://members.gamedev.net/uto/ShadowMap/DepthError.jpg[/link]

aside from using a larger sized texture
u can also look into one of the various non uniform shadowmap generation methods
google should bring quite a few hits
LiSMP (or someit)
PSM
trapezoid shadow map