Ray Tracing Assignment Goal is to reproduce the following So You - PDF document

Ray Tracing Assignment  Goal is to reproduce the following So You Want to Write a Ray Tracer Checkpoint 7 – Tone Reproduction Whitted, 1980 Ray Tracing Assignment Ray Tracing Assignment Seven checkpoints Seven checkpoints    Setting the Scene  Setting the Scene    Camera Modeling  Camera Modeling  Basic Shading  Basic Shading  Procedural Shading  Procedural Shading    Recursive Ray Tracing – Reflection  Recursive Ray Tracing – Reflection  Recursive Ray Tracing – Transmission  Recursive Ray Tracing – Transmission  Tone Reproduction  Tone Reproduction   Checkpoint 7 Tone Reproduction  Goal is to take this from CG units to real units! 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 1

The framework The framework HDR HDR Image Image (10 -8 - 10 8 cd/m 2 ) RESULTS OF RAY TRACER Capture Response Display Capture Response Display compression compression Output Output device device Image Device Image Device Appearance Appearance color management color management The framework Compression HDR  You will implement two different tone Image (10 -8 - 10 8 cd/m 2 ) reproduction operators: RESULTS OF RAY TRACER  Perceptual: Ward’s from Graphics Gems IV Capture Response Display  Photographic: Reinhard, et al. in 2002  Target Device: maximum output luminance of compression new parameter L dmax. Assume standard sRGB Output color space. device Image Device Appearance (0 - Ldmax) color management Color Management The framework  For sake of the checkpoint HDR Image  You can ignore Color Appearance Modeling (10 -8 - 10 8 cd/m 2 ) (assume target viewing conditions is the same as RESULTS OF RAY TRACER the actual viewing condition) Capture Response Display  Assume a simple actual device that has a maximum output of L dmax and a gamma of 1 with standard sRGB color space. compression Output  Simple linear scaling of target image by L dmax device Image Device Appearance (0 - 1) (0 - Ldmax) color management 2

Steps to modifying the ray tracer Step 1: Prepare your HDR image  The values in each image pixel will  New parameters: represent illuminance at that pixel from  L max = maximum illuminance from the the scene. scene  Currently in range of 0-1.  L dmax = maximum display illuminance.  Multiply each channel of each pixel by L max  whichTR = which tone reproduction  Call the result operator to use.  R (x,y) -- illuminance of red channel  G (x,y) -- illuimanance of green channel  B (x,y) -- illuminance of blue channel Step 2: Calculate overall luminance Step 2: Calculate overall illuminance  For both tone reproduction operators,  Like in photography, the operators deal you will need absolute luminance at in luminances and not radiances. each pixel (in addition to luminance on 120 an RGB basis). 100 % Efficiency 80  L (x,y) = absolute luminance at pixel x,y 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 Step 3: Perform compression Step 2: Calculate overall illuminance  Quick and dirty approximation to pixel  You will implement two different tone reproduction operators: illuminance given R, G, B :  Perceptual: Ward’s from Graphics Gems IV  Based on standard CRT color space.  Photographic: Reinhard, et al. in 2002  More on that later. L (x,y) = 0.27 R (x,y)+0.67 G (x,y) +0.06 B (x,y)  Compression:  Note: L is in the range [0, L max ]  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 ] 3

Step 4: Apply device model Questions so far?  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 Let’s talk about the compression (step 3) Ward Tone Reproduction  Operators 2.5 � � sf = 1.219 + ( L d max /2) 0.4  Perceptual: Ward’s from Graphics Gems IV � � 0.4  Photographic: Reinhard, et al. in 2002 1.219 + L wa � �  Ward model: Ward’s TR Operator defines a scale   Simple heuristic based on perceptual tests. factor: L d = sf L w  Includes some visual adaptation.   Easy to implement Where   Gets the job done!!! L wa = adaptation luminance  (can use log-average luminance in scene) Ward Tone Reproduction Log Average Luminance 2.5 � � sf = 1.219 + ( L d max /2) 0.4 � � 0.4  To find the log-average luminance of scene 1.219 + L wa � �  Calculate sf by setting L wa = L � �  L = exp 1 � � log( � + L ( x , y )) �  Final display colors (R target , G target , � � N  � � x , y B target ) are the results of applying  where the sf to the R, G, B .  L (x,y) = absolute luminance at pixel x,y  N = number of pixels  δ = some small number (to prevent log going to infinity) 4

Let’s talk about the compression (step 3) Reinhard Tone Reproduction  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!!!  Mimics Ansel Adam’s Zone System  http://photography.cicada.com/zs/emulator/ Reinhard Tone Reproduction Reinhard Tone Reproduction  Step 1  Basic idea:  Create scaled luminance values R s , G s , B s by mapping  Define the “key” value to the average the key value to Zone V (18% gray) scene luminance and map that to Zone 5. B s = a R s = a G s = a L B ( x , y ) L R ( x , y ) L G ( x , y )  You’ll need log average luminance as caculated for Ward where  Map remaining luminances based on L  = the key value “photographic-like” response.  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., Reinhard Tone Reproduction Reinhard Tone Reproduction  Step 2  Step 3  Calculate target display luminance by simulating illumination.  Find the reflectance for R r , G r , B r , based on film-like  Assume that illuminant is “white” with luminance L dmax response R s ( x , y ) G s ( x , y ) B s ( x , y ) 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 r ( x , y ) = G r ( x , y ) = B r ( x , y ) = R t arg et ( x , y ) = R r ( x , y ) L d max 1 + R s ( x , y ) 1 + G s ( x , y ) 1 + B s ( x , y ) where where  R target , G target , B target (x,y) = target display luminance  R r , G r , B r (x,y) = reflectance (range 0-1)  R s , G s , B s (x,y) = scaled luminance  R r , G r , B r (x,y)= reflectance (range 0-1) 5

Reinhard’s Results 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 Using TR Operator Linear scaling  Please label which is which. Loss of detail [Reinhard,2002] Extra Extra Checkpoint 7  Due dates:  For 5 points  Images to be posted to Web site  Modify the Reinhard operator so that you  Midnight Nov 1st . can specify what luminance is to be used  Final raytracer code to be posted on as the key value. mycourses (dropboxes set up)  Constant value  Midnight Nov 2nd  Value at a pixel  Code must be submitted to receive credit!  Experiment with different illuminant in  Include README with details to build Reinhard model.  Questions? 6