400 likes | 700 Views
High Dynamic Range Imaging. Samu Kemppainen VBM02S. Introduction. Problems with current imaging Basics of HDR images Different techniques Tone mapping HDR compression Image encondings Applications of HDR images. 8-bit problem. Typical images displayed on screen are 24-bits
E N D
High Dynamic Range Imaging Samu Kemppainen VBM02S
Introduction • Problems with current imaging • Basics of HDR images • Different techniques • Tone mapping • HDR compression • Image encondings • Applications of HDR images High Dynamic Range Imaging
8-bit problem • Typical images displayed on screen are 24-bits • 8-bits per color component (RGB) • 256 different intensity levels • Real-world dynamic range is far greater than 256 intensity levels! High Dynamic Range Imaging
Range of luminance in the natural world that the eye can accommodate in a single view 10 000:1 100 000 000:1 that a typical CRT/LCD monitor can display 100:1 (Source: A review of tone reproduction techniques; Devlin, 2002) High Dynamic Range Imaging
What is a High Dynamic Range image? • HDRI is an image that has a greater dynamic range that can be • shown on a display device • captured with a camera with just a single exposure • The "dynamic range" of a scene is the contast ratio between its brightest and darkest parts (Source: http://www.ict.usc.edu/graphics/HDRShop/) High Dynamic Range Imaging
What is a High Dynamic Range image? • Image with a series of images combined into a single high dynamic range image (also called a “radiance map”) • They are useful for representing true illumination values in image-based rendering applications • And for recording incident illumination and using it to illuminate CG objects for realistic composition High Dynamic Range Imaging (Source:http://www.debevec.org/Research/HDR/)
Recovering High Dynamic Range Radiance Maps from Photographs • Acquiring series of differently exposed photographs • Example: sixteen photographs taken at 1-stop increments from 30 seconds to 1/1000 seconds exposure • Then combine the photos by using HDR Shop (Source: Recovering High Dynamic Range Radiance Maps from Photographs; Debevec & Malik, 1997) High Dynamic Range Imaging
Recovering HDR radiance maps from photographs Sixteen photographs taken at 1-stop increments from 30 seconds to 1/1000 seconds exposure (Source: Recovering High Dynamic Range Radiance Maps from Photographs; Debevec & Malik, 1997) High Dynamic Range Imaging
Low dynamic range -vs- High dynamic range Original photograph (LDR) Motion-blurred LDR image Motion-blurred HDR image Photograph with real motion blur High Dynamic Range Imaging (Source:http://www.debevec.org/Research/HDR/)
How to view HDR imageson a LDR device? • By using Real-Time High Dynamic Range Texture Mapping • hardware pipeline • Using Gradient Domain High Dynamic Range Compression (Source:Real-Time High Dynamic Range Texture Mapping; Cohen et al., 2001) High Dynamic Range Imaging
High Dynamic Range Texture Mapping • HDR textures are decomposed into sets of 8-bit textures • They are dynamically reassembled by the graphics hardware’s multitexturing system • or using multipass techniques and framebuffer image prosessing • Only two 8-bit textures is needed to be in memory simultaneously (Source:Real-Time High Dynamic Range Texture Mapping; Cohen et al., 2001) High Dynamic Range Imaging
HDR Texture Mapping pipeline (Source:Real-Time High Dynamic Range Texture Mapping; Cohen et al., 2001) High Dynamic Range Imaging
Tone mapping • Tone Reproduction and Physically Based Spectral Rendering • Develope for use in tv and photography • Takes advantage of HVS (Human Visual System): the fact that HVS has a greater sensitivity torelative rather than absolute luminance levels (Source: Tone Reproduction and Physically Based Spectral Rendering; Devlin et al., 2002) High Dynamic Range Imaging
(Dynamic range: 25 000:1) High Dynamic Range Imaging (Source: http://www.cs.huji.ac.il/~danix/hdr/)
Larson et al., 1997 Low Curvature Image Simplifier (LCIS) method (Tumblin & Turk, 1999) High Dynamic Range Imaging (Source: http://www.cs.huji.ac.il/~danix/hdr/)
Gradient Domain High Dynamic Range Compression • Way of rendering HDR images on conventional display • technique is effective, fast and simple • less artifacts than with other methods • it also significantly enhance ordinary images by bringing out detail in dark regions (Source: Gradient Domain High Dynamic Range Compression; Fattal et al., 2002) High Dynamic Range Imaging
Gradient Domain HDR Compression • Based on the simple idea • identify large gradients at various scale • attenuate the their magnitudes progressively • a reduced HDR image is then recontructed from the attenuated gradient field (Source: Gradient Domain High Dynamic Range Compression; Fattal et al., 2002) High Dynamic Range Imaging
The Gradient attenuation factors used to compress HDR radiance map High Dynamic Range Imaging (Source: http://www.cs.huji.ac.il/~danix/hdr/)
Gradient Domain HDR Compression High Dynamic Range Imaging (Source: http://www.cs.huji.ac.il/~danix/hdr/)
Example using only two images in Gradient Domain HDR Compression 1/22 seconds, f8 1 second, f8 High Dynamic Range Imaging (Source: http://www.cs.huji.ac.il/~danix/hdr/)
High dynamic range image encodings • HDR image have to be encoded to encompass a large range of values • Some file formats: • Pixar Log Encoding (TIFF) • Radiance RGBE Encoding (HDR) • SGI LogLuv (TIFF) • ILM OpenEXR (EXR) • Microsoft/HP scRGB Encoding High Dynamic Range Imaging (Source: http://www.anyhere.com/gward/hdrenc/)
Cost (bits/pixel) -vs- benefit (dynamic range) of full-gamut formats High Dynamic Range Imaging (Source: http://www.anyhere.com/gward/hdrenc/)
The difference in the encodings • Benefits of log and floating point representations over linear or gamma encodings • 24 bits LogLuv format holds more dynamic range than the 36-bit scRGB-nl format (gamma encoding) and even the 48-bit scRGB linear encoding • 32 bits LogLuv encoding holds 10 times the dynamic range over the scRGB and scRGB-nl. • The EXR encoding holds 3 times the range of scRGB encoding in the same 48 bits, with much higher precision than any of the other formats High Dynamic Range Imaging (Source: http://www.anyhere.com/gward/hdrenc/)
Image results • Some rounding error occur when using log or floating points • ΔE*=2: Detectable under ideal conditions • ΔE*=5: Noticeable in side-by-side images High Dynamic Range Imaging
False color difference view of 24-bit LogLuv High Dynamic Range Imaging (Source: http://www.anyhere.com/gward/hdrenc/)
24-bit LogLuv 32-bit LogLuv EXR RGBE XYZE scRGB scRGB-nl scYCC-nl High Dynamic Range Imaging (Source: http://www.anyhere.com/gward/hdrenc/)
Encoding quality curves (average) High Dynamic Range Imaging (Source: http://www.anyhere.com/gward/hdrenc/)
Some applications of High Dynamic Range images • Global illumination techniques (physically-based rendering) • Mixed reality rendering (special effects for movies and commercials) • Human vision simulation and psychophysics • Reconnaissance and satellite imaging (remote sensing) • Digital compositing for film • Digital cinema High Dynamic Range Imaging
Thank you for listening! Questions? Comments? High Dynamic Range Imaging