Adaptive ratio images in HDR image representation
Image encoding and decoding are described. An input HDR image is coded using a coded tone-mapped image, a coded invertible function of a ratio image, and a look-up table representing the inverse of the invertible function, wherein the ratio image represents the pixel by pixel ratio of the luminance pixel values of the input image and the corresponding luminance values in the tone-mapped image. A method to generate the invertible function based on a cumulative histogram of the ratio image is presented.
1. A method to code a high dynamic range (HDR) image with a processor, the method comprising:
receiving an input image with a first dynamic range;
generating a tone-mapped image based on the input image, wherein the tone-mapped image has a dynamic range lower than the first dynamic range;
generating a ratio image (Y R ), the ratio image comprising luminance pixel values of the input image divided pixel by pixel by corresponding luminance pixel values in the tone-mapped image;
applying an invertible function (F) to the ratio image to generate a modified ratio image (F(Y R ));
generating a look-up table to represent the inverse of the invertible function (F −1 ), wherein applying the look-up table (F −1 ) to the modified ratio image generates an approximation of the ratio image; and
generating a coded HDR image based on the tone-mapped image and the modified ratio image, wherein the invertible function is selected so that when applying its inverse to the modified ratio image, the distance between the ratio image and the approximation of the ratio image is minimized according to a predetermined measure.
2. The method of claim 1 , wherein the distance between the ratio image and the approximation of the ratio image is minimized according to a mean square error criterion.
3. The method of claim 1 , wherein the predetermined measure comprises the signal to noise ratio between the ratio image and the approximation of the ratio image.
4. The method of claim 1 , wherein the predetermined measure comprises the peak signal to noise ratio of the approximation of the ratio image.
5. The method of claim 1 , wherein a file format of the coded HDR image comprises a coded tone-mapped image, a coded version of the modified ratio image, and the look-up table representing the inverse of the invertible function.
6. The method of claim 5 , wherein, in a decoder, decoding the coded HDR image comprises:
receiving the coded HDR image;
extracting the tone-mapped image, the modified ratio image, and the look-up table from the coded HDR image;
applying the look-up table to the modified ratio image to generate the approximation of the ratio image; and
generating an output HDR image based on the tone-mapped image and the approximation of the ratio image.
7. The method of claim 1 , further comprising quantizing the modified ratio image to generate a quantized ratio image.
8. The method of claim 7 , wherein quantizing the modified ratio image comprises computing
Y
Q
=
round
(
scale
*
F
(
Y
R
)
-
min
(
F
(
Y
R
)
)
max
(
F
(
Y
R
)
)
-
min
(
F
(
Y
R
)
)
)
,
wherein Y Q denotes the quantized ratio image, Y R denotes the ratio image, F(Y R ) denotes the modified ratio image, scale denotes a scaling factor, and min(F(Y R )) and max(F(Y R )) denote the minimum and maximum values of the modified ratio image.
9. The method of claim 8 , wherein scale=2 B −1 where B denotes the number of bits for representing a pixel of the quantized ratio image.
10. The method of claim 1 , further comprising:
computing a histogram of pixel values of the ratio image;
computing a cumulative histogram based on the histogram; and
generating the invertible function based on the cumulative histogram.
11. The method of claim 10 , wherein generating the invertible function comprises computing
F i =(( c _hist i −min( c _hist))/(max( c _hist)−min( c _hist))*scale,
wherein c_hist i denotes the cumulative histogram value for the i-th pixel value of the ratio image, min(c_hist) and max(c_hist) denote the minimum and maximum values of the cumulative histogram, and scale denotes a scaling factor.
12. The method of claim 11 , wherein scale=2 B −1, where B denotes the number of bits for representing a pixel of the modified ratio image.
13. A non-transitory computer-readable storage medium having stored thereon computer-executable instructions for executing with a processor a method in accordance with claim 1 .
14. An apparatus to decode a coded high-dynamic range (HDR) image, the apparatus comprising:
a processor; and
a non-transitory computer-readable storage medium having stored thereon computer-executable instructions for executing with the processor a method in accordance with claim 1 .