Ray Tracing Assignment Goal is to reproduce the following Whitted, - PDF document

So You Want to Write a Ray Tracer Checkpoint 7 – Tone Reproduction Ray Tracing Assignment  Goal is to reproduce the following Whitted, 1980 1

Ray Tracing Assignment Seven checkpoints   Setting the Scene   Camera Modeling  Basic Shading  Procedural Shading   Recursive Ray Tracing – Reflection  Recursive Ray Tracing – Transmission  Tone Reproduction  Ray Tracing Assignment Seven checkpoints   Setting the Scene   Camera Modeling  Basic Shading  Procedural Shading   Recursive Ray Tracing – Reflection  Recursive Ray Tracing – Transmission  Tone Reproduction  2

Checkpoint 7  Goal is to take this from CG units to real units! Tone Reproduction Change your ray tracer so that it:   Maps lighting units (0-1) to real lighting  units  Applies a tone reproduction operator to compress these simulated radiances to display radiances 3

The framework HDR Image Capture Response Display compression Output device Image Device Appearance color management The framework HDR Image (10 -8 - 10 8 cd/m 2 ) RESULTS OF RAY TRACER Capture Response Display compression Output device Image Device Appearance color management 4

Compression  You will implement two different tone reproduction operators:  Perceptual: Ward’s from Graphics Gems IV  Photographic: Reinhard, et al. in 2002  Target Device: maximum output luminance of new parameter L dmax. Assume standard sRGB color space. The framework HDR Image (10 -8 - 10 8 cd/m 2 ) RESULTS OF RAY TRACER Capture Response Display compression Output device Image Device Appearance (0 - Ldmax) color management 5

Color Management  For sake of the checkpoint  You can ignore Color Appearance Modeling (assume target viewing conditions is the same as the actual viewing condition)  Assume a simple actual device that has a maximum output of L dmax and a gamma of 1 with standard sRGB color space.  Simple linear scaling of target image by L dmax The framework HDR Image (10 -8 - 10 8 cd/m 2 ) RESULTS OF RAY TRACER Capture Response Display compression Output device Image Device Appearance (0 - 1) (0 - Ldmax) color management 6

Steps to modifying the ray tracer  New parameters:  L max = maximum illuminance from the scene  L dmax = maximum display illuminance.  whichTR = which tone reproduction operator to use. Step 1: Prepare your HDR image  The values in each image pixel will represent illuminance at that pixel from the scene.  Currently in range of 0-1.  Multiply each channel of each pixel by L max  Call the result  R (x,y) -- illuminance of red channel  G (x,y) -- illuimanance of green channel  B (x,y) -- illuminance of blue channel 7

Step 2: Calculate overall luminance  For both tone reproduction operators, you will need absolute luminance at each pixel (in addition to luminance on an RGB basis).  L (x,y) = absolute luminance at pixel x,y Step 2: Calculate overall illuminance  Like in photography, the operators deal in luminances and not radiances. 120 100 % Efficiency 80 60 40 20 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 5 0 7 0 2 5 7 0 2 5 7 0 2 5 7 0 2 5 3 4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 Wavelength 8

Step 2: Calculate overall illuminance  Quick and dirty approximation to pixel illuminance given R, G, B :  Based on standard CRT color space. L (x,y) = 0.27 R (x,y)+0.67 G (x,y) +0.06 B (x,y)  Note: L is in the range [0, L max ] Step 3: Perform compression  You will implement two different tone reproduction operators:  Perceptual: Ward’s from Graphics Gems IV  Photographic: Reinhard, et al. in 2002  More on that later.  Compression:  will be applied on R(x,y), G(x,y), B(x,y)  defines R target (x,y), G target (x,y), B target (x,y)  In range [0, L dmax ] 9

Step 4: Apply device model  Assume a simple actual device that has a maximum output of L dmax and a gamma of 1 with standard sRGB color space.  Simple linear scaling of target image by L dmax  R final (x,y) = R target (x,y) / L dmax  G final (x,y) = G target (x,y) / L dmax  B final (x,y) = B target (x,y) / L dmax Questions so far? 10

Let’s talk about the compression (step 3)  Operators  Perceptual: Ward’s from Graphics Gems IV  Photographic: Reinhard, et al. in 2002  Ward model:  Simple heuristic based on perceptual tests.  Includes some visual adaptation.  Easy to implement  Gets the job done!!! Ward Tone Reproduction 2.5 � � sf = 1.219 + ( L d max /2) 0.4 � � 0.4 1.219 + L wa � � Ward’s TR Operator defines a scale  factor: L d = sf L w  Where  L wa = adaptation luminance  (can use log-average luminance in scene) 11

Log Average Luminance  To find the log-average luminance of scene � � L = exp 1 � log( � + L ( x , y )) � � � � N � � x , y  where  L (x,y) = absolute luminance at pixel x,y  N = number of pixels  δ = some small number (to prevent log going to infinity) Ward Tone Reproduction 2.5 � � sf = 1.219 + ( L d max /2) 0.4 � � 0.4 1.219 + L wa � �  Calculate sf by setting L wa = L   Final display colors (R target , G target , B target ) are the results of applying the sf to the R, G, B . 12

Let’s talk about the compression (step 3)  Operators  Perceptual: Ward’s from Graphics Gems IV  Photographic: Reinhard, et al. in 2002  Reinhard model:  Simple heuristic based on photographic systems.  Models photographic-like response.  Easy to implement  Gets the job done!!! Reinhard Tone Reproduction  Mimics Ansel Adam’s Zone System  http://photography.cicada.com/zs/emulator/ 13

Reinhard Tone Reproduction  Basic idea:  Define the “key” value to the average scene luminance and map that to Zone 5.  You’ll need log average luminance as caculated for Ward  Map remaining luminances based on “photographic-like” response. Reinhard Tone Reproduction  Step 1  Create scaled luminance values R s , G s , B s by mapping the key value to Zone V (18% gray) B s = a R s = a G s = a L B ( x , y ) L R ( x , y ) L G ( x , y ) where L  = the key value  R/G/B(x,y) = scene luminance at pixel x,y  a = % gray for zone V; use a = 0.18  R s , G s , B s (x,y) = scaled luminance, i.e., 14

Reinhard Tone Reproduction  Step 2  Find the reflectance for R r , G r , B r , based on film-like response R s ( x , y ) G s ( x , y ) B s ( x , y ) R r ( x , y ) = G r ( x , y ) = B r ( x , y ) = 1 + R s ( x , y ) 1 + G s ( x , y ) 1 + B s ( x , y ) where  R s , G s , B s (x,y) = scaled luminance  R r , G r , B r (x,y)= reflectance (range 0-1) Reinhard Tone Reproduction  Step 3  Calculate target display luminance by simulating illumination.  Assume that illuminant is “white” with luminance L dmax G t arg et ( x , y ) = G r ( x , y ) L d max B t arg et ( x , y ) = B r ( x , y ) L d max R t arg et ( x , y ) = R r ( x , y ) L d max where  R target , G target , B target (x,y) = target display luminance  R r , G r , B r (x,y) = reflectance (range 0-1) 15

Reinhard’s Results Using TR Operator Linear scaling Loss of detail [Reinhard,2002] Checkpoint 7  To be posted to Web site  Six images produced by running your raytracer with three different values of L max for EACH tone reproduction operator:  Lo-Range Lighting: L max = 1 nit  Mid-Range Lighting: L max = 1000 nits  Hi-Range Lighting: L max = 10,000 nits  Tone Reproduction:  Ward’s Model  Reinhard’s Model  Please label which is which. 16

Extra Extra  For 5 points  Modify the Reinhard operator so that you can specify what luminance is to be used as the key value.  Constant value  Value at a pixel  Implement vignetting. Checkpoint 7  Due dates:  Images to be posted to Web site  Wednesday, Feb 13th .  Final raytracer code to be posted on mycourses (dropboxes set up)  Feb 15th.  Code must be submitted to receive credit!  Include README with details to build  Questions? 17