Display of

High-Dynamic-Range Images
(Tone mapping)
EE 5176: Computational Photography
High-dynamic-range (HDR) images
• CG Images

• Multiple exposure photo [Debevec & Malik 1997]

Recover
HDR value
response
for each pixel
curve

• HDR sensors
 Contrast reduction
• Match limited contrast of the medium
• Preserve details

10-6 High dynamic range 106

Real world

10-6 106
Picture

Low contrast
Overview
➢Global tone mapping
➢ Gamma correction
➢ Photographic tone mapping

➢Fast bilateral filtering for HDR display (local

tone mapping)
➢ Motivation
➢ Review of bilateral filtering
➢ Acceleration
➢ Use for contrast reduction
Overview
➢Global tone mapping
➢ Gamma correction
➢ Photographic tone mapping

➢Fast bilateral filtering for HDR display (local

tone mapping)
➢ Motivation
➢ Review of bilateral filtering
➢ Acceleration
➢ Use for contrast reduction
 Linear mapping does not work

Linear mapping Better non-linear mapping

A(I/max(I)), where A is dynamic range of
display such as 255
Gamma correction

Simple non-linear mapping

Iout= A(I/max(I))ɤ
A is the dynamic range of display such as 255

Image coutesy wikipedia

 Photographic tone reproduction
Reinhard et al 2002
The key of a scene indicates whether it is subjectively light, normal, or dark. A white-painted
room would be high-key, and a dim stable would be low-key.

Log average is useful

approx. to key value of the
scene.

For scene with

normal-key, a=0.18.
Photographic tone reproduction
➢ Tone mapping
➢ Range should be 0-1
➢ Should compress large intensities more
➢ Large intensities scaled by 1/L
➢ Small intensities scaled by 1

➢ To saturate high intensities, use

➢ where Lwhite is the smallest value that will be mapped to 1.

Photographic tone reproduction

Linear mapping Photographic tone mapping

Overview
➢Global tone mapping
➢ Gamma correction
➢ Photographic tone mapping

➢Fast bilateral filtering for HDR display (local

tone mapping)
➢ Motivation
➢ Review of bilateral filtering
➢ Acceleration
➢ Use for contrast reduction
 Fast Bilateral Filtering
for the Display of
High-Dynamic-Range Images

Frédo Durand & Julie Dorsey

Laboratory for Computer Science
Massachusetts Institute of Technology
A typical photo
• Sun is overexposed
• Foreground is underexposed
 Gamma compression
• X -> X g
• Colors are washed-out
Input Gamma
 Gamma compression on intensity
• Colors are OK,
but details (intensity high-frequency) are blurred
Intensity Gamma on intensity

Color
 Chiu et al. 1993
• Reduce contrast of low-frequencies
• Keep high frequencies
Low-freq. Reduce low frequency

High-freq.

Color
 The halo nightmare
• For strong edges
• Because they contain high frequency
Low-freq. Reduce low frequency

High-freq.

Color
Figure courtesy Tumblin 99
Figure courtesy Tumblin 99
 Our approach
• Do not blur across edges
• Non-linear filtering
Large-scale Output

Detail

Color
Edge-preserving filtering
• Blur, but not across edges
Input Gaussian blur Edge-preserving

• Anisotropic diffusion [Perona & Malik 90]

– Blurring as heat flow
– LCIS [Tumblin & Turk]
• Bilateral filtering [Tomasi & Manduci, 98]
Plan
• Review of bilateral filtering [Tomasi and Manduchi 1998]

• Acceleration

• Use for contrast reduction

Start with Gaussian filtering
• Here, input is a step function + noise

J = f  I

output input
Start with Gaussian filtering
• Spatial Gaussian f

J = f  I

output input
Start with Gaussian filtering
• Output is blurred

J = f  I

output input
Gaussian filter as weighted average
• Weight of x depends on distance to x

J (x) = x f ( x, x ) I (x )

x
x x

output input
The problem of edges
• Here, I (x ) “pollutes” our estimate J(x)
• It is too different

J (x) = x f ( x, x ) I (x )

I (x)
x
I (x )

output input
Principle of Bilateral filtering
[Tomasi and Manduchi 1998]

• Penalty g on the intensity difference

1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )

I (x)
x
I (x )

output input
Bilateral filtering
[Tomasi and Manduchi 1998]

• Spatial Gaussian f

1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )

output input
Bilateral filtering
[Tomasi and Manduchi 1998]

• Spatial Gaussian f
• Gaussian g on the intensity difference
1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )

output input
Normalization factor
[Tomasi and Manduchi 1998]

• k(x)= x f ( x, x ) g ( I (x ) - I ( x))

1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )

output input
Bilateral filtering is non-linear
[Tomasi and Manduchi 1998]

• The weights are different for each output pixel

1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )

x x
output input
Plan
• Review of bilateral filtering [Tomasi and Manduchi 1998]

• Acceleration

• Use for contrast reduction

Acceleration
• Non-linear because of g

1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )
Acceleration
• Linear for a given value of I(x)
• Convolution of g I by Gaussian f

1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )
Acceleration
• Linear for a given value of I(x)
• Convolution of g I by Gaussian f
• Valid for all x with same value I(x)
1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )
Acceleration
• Discretize the set of possible I(x)
• Perform linear Gaussian blur (FFT)
• Linear interpolation in between
1
J (x) =
k ( x)
x f ( x, x ) g ( I (x ) - I ( x)) I (x )

• k(x) treated similarly

k(x)=  f ( x, x ) g ( I (x ) - I ( x))
Plan
• Review of bilateral filtering [Tomasi and Manduchi 1998]

• Acceleration

• Use for contrast reduction

 Contrast reduction
Input HDR image

Contrast
too high!
 Contrast reduction
Input HDR image

Intensity

Color
 Contrast reduction
Input HDR image

Intensity Large scale

Fast
Bilateral
Filter

Color
 Contrast reduction
Input HDR image

Intensity Large scale

Fast Detail
Bilateral
Filter

Color
 Contrast reduction
Input HDR image

Intensity Large scale Large scale

Reduce
contrast

Fast Detail
Bilateral
Filter

Color
 Contrast reduction
Input HDR image

Intensity Large scale Large scale

Reduce
contrast

Fast Detail Detail

Bilateral Preserve!
Filter

Color
 Contrast reduction
Input HDR image Output

Intensity Large scale Large scale

Reduce
contrast

Fast Detail Detail

Bilateral Preserve!
Filter

Color Color
Algorithm details

• Input intensity= 1/61*(R*20+G*40+B)

• r=R/(input intensity), g=G/input intensity, B=B/input intensity
• log(base)=Bilateral(log(input intensity))
• log(detail)=log(input intensity)-log(base)
• log (output intensity)=log(base)*compressionfactor+log(detail) -
log_absolute_scale
• R output = r*10^(log(output intensity)).

See https://people.csail.mit.edu/fredo/PUBLI/Siggraph2002/
Informal comparison

Gradient-space Bilateral Photographic

[Fattal et al.] [Durand et al.] [Reinhard et al.]
Informal comparison

Gradient-space Bilateral Photographic

[Fattal et al.] [Durand et al.] [Reinhard et al.]
Overview
➢Global tone mapping
➢ Gamma correction
➢ Photographic tone mapping

➢Fast bilateral filtering for HDR display (local

tone mapping)
➢ Motivation
➢ Review of bilateral filtering
➢ Acceleration
➢ Use for contrast reduction