Image Tone Mapping for HDR Images with Edge Preserving Filtering

HDR (High Dynamic Ranging) Imagery

• • A starlit night has an average luminance level of around 10^-3 candelas/m^2, and daylight scenes are close to 10^5 cd/m^2 ;

  • Humans can see detail in regions that vary by 1:10^4 at any given adaptation level, over which the eye gets swamped by stray light (i.e., disability glare) and details are lost;

  • A well-designed CRT monitor's maximum display luminance is only around 100 cd/m^2;

  • A high-quality xenon film projector is still two orders of magnitude away from the optimal light level for human acuity and colour perception;

  • HDR image save each pixel with 4 bytes (32bits);

  • Two main methods for generating HDR imaging:

    • • Use physically based renderers, generating basically all visible colours;

      • Take photographs of a particular scene at different exposure times, apply radiometric calibration to each camera, and combine the calibrated images.

Fig. 1, Range of luminances in the real world compared to RGB

Fig. 2, Taking pictures at multiple exposure times and camera response function

Tone Mapping (Tone Reproduction)

• • Match one observer model applied to the scene to another observer model applied to the desired display image;

  • HDR needs tone mapping to display colors on monitor;

    • Tone Mapping operators' 4 Categories

      • global (spatially uniform): Compress images using an identical (non-linear) curve for each pixel;

      • local (spatially varying): Achieve dynamic range reduction by modulating a non-linear curve for each pixel independently (local neighborhood);

      • frequency domain: Reduce the dynamic range of image components selectively, based on their spatial frequency;

      • gradient domain: Modify the derivative of an image to achieve dynamic range reduction. (Gradient domain HDR compression);

    • Video tone mapping: temporal coherence;

    • Inverse tone mapping: from LDR to HDR;

Edge Aware/Preserving Filter:

• • TV (Total Variation) for edge preserving filtering;

  • Anistropic diffusion (robust);

  • Bilateral filter;

  • Weighted LS method;

  • EAW (Edge Avoiding Wavelet) filter;

  • Non-local means;

  • L0 Gradient minimization;

  • Domain Transform for edge preserving filtering;

  • Local Laplacian Filters for edge preserving processing;

  • BM3D (block matching 3-d collaborative filtering).

Image Decomposition

• • Detail.img = Input.img/Base.img: I(x,y) = R(x,y)*L(x,y);

  • Smooth base layer (large scale variations);

  • Residual detail layer (small scale details);

Tone Mapping based on Image Decomposition: A Frequency Domain Method

• • Compression in illumination layer;

  • Unchanged in reflectance layers.

Fig. 3, Tone mapping results for HDR images (left: bilaterl, middle: WLS, right: bm3d)

References

1. R. Fattal, Gradient domain HDR compression, Siggraph, 2002.

2. Z. Farbman et al., Edge-preserving decompositions for multi-scale tone and detail manipulation, Siggraph, 2008.

3. E. Gastal, M. Oliveira, Domain transform for edge-aware image and video processing, ACM SIGGRAPH 2011.

4. L. Xu et al., Image smoothing via l0 gradient minimization, ACM SIGGRAPH Asia 2011.

5. H.T. Chen, Tone Reproduction: A Perspective from Luminance-Driven Perceptual Grouping, International Journal of Computer Vision, 2005.

6. L. Meylan, High Dynamic Range Image Rendering With a Retinex-Based Adaptive Filter, IEEE Transactions on Image Processing, 2006.

7. Raanan Fattal, Edge-Avoiding Wavelets and their Applications, ACM Trans. Graph., 28(3), 2009.

8. Sylvain Paris, Samuel W. Hasinoff, Jan Kautz, Local Laplacian Filters: Edge-aware Image Processing with a Laplacian Pyramid , ACM SIGGRAPH, 2011.

9. E. Reinhard, Photographic Tone Reproduction for Digital Images, Proceedings of SIGGRAPH 2002.

10. X.G. Li, An Adaptive Algorithm for the Display of High Dynamic Range Images, Journal of Visual Communication and Image Representation, 2007.