Rendering with Spherical Harmonics - computer graphics research随着计算机图形绘制球面谐波研究.pptVIP

Rendering with Spherical Harmonics Mark ColbertPhD CandidateUCF Graphics Group / MCL colbert@cs.ucf.edu welcome Previously GPU Pipeline OpenGL/Cg API DirectX API Questions? overview Orthogonal Basis Functions Spherical Harmonics Illumination Rotation Reparameterization orthogonal basis functions Space to represent data Different spaces often allow for compression of coefficients Lets look at one simple example of the following piece of data orthogonal basis functions Standard Basis DCT Discrete Cosine Transform Use Cosine waves as basis functions building our data set with DCT Contrived Case building our data set with DCT Only needed 3 coefficients instead of 20! Remaining coefficients all 0 Most of the time data not perfect Obtain good reconstruction from few coefficients Arbitrary function conversion requires projection orthogonal basis functions Property Integration of two functions is equivalent to scalar product of coefficients spherical harmonics We only will look at real (non-complex) spherical harmonics Where l = 0…?, m=-l…l P is the Associated Legendre Polynomial spherical harmonics Demo Converting to Cartesian Coordinates illumination Diffuse Reflections Special (but common) Case Outgoing Radiance Is Not Viewing Dependent illumination We represent the irradiance map for a given position as Where R is the rotation of the incoming light by some angle, alpha beta coefficients Projection Inverse rotation reparameterization Rotatable Major advantage over other popular basis functions such as Wavelets Idea Most Ideal BRDFs have an 1D falloff along a given axis Lambertian – Normal Vector Phong – Reflection Vector Blinn – Half-angle Vector rotation reparameterization Reparameterization example Lambertian BRDF Only need first 2 band for most of the energy! (9 SH coefficients) rotation reparameterization For this case No azimuthal variance Only zonal coefficients needed illumination Payoff illumination If we want to define rotation in terms of t