REVECTORIZATION-BASED SHADOW MAPPING

Transcription

1 REVECTORIZATION-BASED SHADOW MAPPING Márcio C. F. Macedo (UFBA Brazil) Antônio L. Apolinário Jr. (UFBA Brazil) PGCOMP (UFBA Brazil)

2 AGENDA Introduction; Revectorization-Based Shadow Mapping: Revectorization Pipeline; Single-Pass Shadow Map Silhouette Revectorization; Revectorization-Based Shadow Map Silhouette Smoothing; Revectorization-Based Percentage-Closer Filtering; Results and Discussion; Conclusion and Future Work; 2

3 3 INTRODUCTION

4 CONTEXT No Shadow Shadow Mapping 4

5 CURRENT SCENARIO Filtering Silhouette Recovery 5 Low Computational Cost High Computational Cost

6 CONTRIBUTIONS We introduce the Revectorization-Based Shadow Mapping (RBSM) for real-time rendering of highquality, anti-aliased hard shadows; Three methods are presented: Shadow Map Silhouette Revectorization (SMSR); Revectorization-Based Shadow Map Silhouette Smoothing (RSMSS); Revectorization-Based Percentage-Closer Filtering (RPCF); 6

7 REVECTORIZATION-BASED SHADOW MAPPING 7

8 Overview: REVECTORIZATION PIPELINE SMSR Shadow Map Entering Discontinuity Entering ONDS Clipped ONDS Shadowed Scene Entering + Exiting Discontinuities Entering + Exiting ONDS Smoothed ONDS Shadowed Scene Shadow Test RSMSS 8

9 REVECTORIZATION PIPELINE Discontinuity: Entering; Exiting; Jagged Shadow Edge Discontinuity Space 9

10 REVECTORIZATION PIPELINE Oriented Normalized Discontinuity Space (ONDS): Entering Discontinuity Exiting Discontinuity Entering Discontinuity Entering ONDS Entering + Exiting Discontinuities Entering + Exiting ONDS 10

11 SHADOW MAP SILHOUETTE REVECTORIZATION (SMSR) Visibility Function: Deal with 12 different shadowing configurations; 11

12 SHADOW MAP SILHOUETTE REVECTORIZATION (SMSR) Shadow Mapping SMSR 12

13 REVECTORIZATION-BASED SHADOW MAP SILHOUETTE SMOOTHING (RSMSS) Visibility Function: Deal with 31 different shadowing configurations; 13

14 REVECTORIZATION-BASED SHADOW MAP SILHOUETTE SMOOTHING (RSMSS) Shadow Mapping RSMSS 14

15 REVECTORIZATION-BASED PERCENTAGE- CLOSER FILTERING (RPCF) Goal: Enable control over the filter size; Algorithm: Evaluate the revectorization-based visibility function for every texel inside the RPCF kernel; Compute the averaged final shadow intensity; 15

16 REVECTORIZATION-BASED PERCENTAGE- CLOSER FILTERING (RPCF) Shadow Mapping RPCF + RSMSS 16

17 17 RESULTS AND DISCUSSION

18 EXPERIMENTAL SETUP For all tests we used an Intel Core i7-3770k 8GB RAM, NVIDIA GeForce GTX 660; Memory requirements were computed considering the mip-map overhead by a factor 1.3; Filtering techniques were tested using a 3x3 box filter; 18

19 Temporal Coherence: RENDERING QUALITY 19

20 Silhouette Recovery: RENDERING QUALITY Shadow Mapping SMSR Shadow Volumes 20

21 Silhouette Recovery: PERFORMANCE 21

22 RENDERING QUALITY Filtering: Exponential Variance Moment RPCF Shadow Variance Shadow RSMSS PCF + SMSR RSMSS Shadow Mapping Mapping 22

23 PERFORMANCE Filtering: PCF (3x3) PCF (17x17) RPCF + RSMSS (3x3) 23

24 MEMORY CONSUMPTION Memory Consumption (MB) ² 1024² 2048² 4096² 50 0 SM PCF RBSM VSM ESM EVSM MSM Technique 24

25 RENDERING QUALITY Fine Details + Complex Crossing Edges: 25

26 CONCLUSION AND FUTURE WORK 26

27 FINAL CONSIDERATIONS Conclusion: Our techniques are accurate, provide consistent realtime frame rates, and do not suffer from light leaking artifacts; Our approach is useful for games and other interactive applications; We believe that the RPCF is a good alternative to the PCF algorithm; Future Work: Extend the use of revectorization for soft shadows; Improve accuracy by incorporating additional geometric information into the revectorization pipeline; 27

28 We are grateful to: ACKNOWLEDGMENTS Vladimir Bondarev - for discussing the ideas behind his original implementation of the two-pass SMSR method; CAPES for financial support; 28

29 Thank You! Márcio C. F. Macedo Antônio L. Apolinário Jr. 29