Block-based content-adaptive reshaping for high dynamic range images

Patent metadata

Field	Value
Publication number	US-10032262-B2
Application number	US-201715418199-A
Country	US
Kind code	B2
Filing date	Jan 27, 2017
Priority date	Feb 2, 2016
Publication date	Jul 24, 2018
Grant date	Jul 24, 2018

Abstract

A processor for signal reshaping receives an input image with an input bit depth. Block-based standard deviations are computed. The input codewords are divided into codeword bins and each bin is assigned a standard deviation value. For each bin, a standard deviation to bit-depth function is applied to the bin values to generate minimal bit depth values for each codeword bin. An output codeword mapping function is generated based on the input bit depth, a target bit depth, and the minimal bit depth values. The codeword mapping function is applied to the input image to generate an output image in the target bit depth.

First claim

What is claimed is:

1. A method for adaptive image reshaping with a processor, the method comprising:
accessing with a processor an input image in an input bit depth;
dividing the input image into a plurality of blocks and computing for one or more image blocks a measure of complexity of their pixel values;
dividing a range of input codewords into a plurality of codeword bins;
for one or more of the codeword bins:
computing a bin measure of complexity value based on the block measures of complexity of at least the input image;
applying a measure of complexity to bit-depth function to generate a minimal bit-depth value;

generating a codeword mapping function based on the input bit depth, the minimal bit depth values, and a target bit depth, wherein the codeword mapping function maps input codewords in the input bit depth to output codewords in the target bit depth and wherein generating the codeword mapping function further comprises:
for each codeword bin:
generating a lower bound of normalized number of required codewords in the target bit depth based on the minimal bit depth values, the input bit depth and the target bit depth;

allocating unused normalized codewords to each bin image according to an allocation scheme to generate updated normalized numbers of required codewords; and
generating the codeword mapping function by computing a cumulative sum of the updated normalized numbers of required codewords, wherein generating the codeword mapping function for an input pixel value i comprises computing

FL
⁡

(
i
)

=

∑

k
=
0

i

⁢

s
k

,

wherein s k values are derived based on the updated normalized number of codeword values; and
applying the codeword mapping function to the input image to generate an output image in the target bit depth.

2. The method of claim 1 , wherein computing the measure of complexity of pixel values in a block comprises computing the standard deviation of the pixel values in the block.

3. The method of claim 1 , further comprising:
applying a pixel selection process to the input image to eliminate computing block measures of complexity for pixels in a letterbox area of the input image.

4. The method of claim 1 wherein the input image is a high dynamic range image encoded according to gamma encoding or SMPTE ST 2084.

5. The method of claim 1 , wherein the measure of complexity to bit-depth function is generated according to results from a perceptual user study.

6. The method of claim 5 , wherein the perceptual user study comprises:
accessing a plurality of original high dynamic range (HDR) images in the input bit depth;
for each original image in the plurality of high dynamic range images:
converting the original image from its original color space to a second color space;
truncating the image in the second color space to generate truncated images at bit depths lower than the input bit depth;
converting the truncated images to the original color space to generate reconstructed images; and
determining the lower bit-depth for which one of the truncated images in the original color space best matches the original image.

7. The method of claim 1 , further comprising, filtering the updated normalized numbers of required codewords by a low-pass filter before generating the codeword mapping function.

8. The method of claim 1 , wherein the allocation scheme comprises a constant offset allocation scheme, where for the i-th input pixel value

d
~

i

=

d
i

+

U

v
H

-

v
L

,

wherein {tilde over (d)} i denotes the updated normalized number of codeword values, d i denotes the normalized number of required codewords, if D denotes the sum of d i values, then U=1−D, v H denotes a maximum input pixel value, and v L denotes a minimum input pixel value.

9. The method of claim 1 , wherein the allocation scheme comprises computing
{tilde over (d)} i =d i +( u i −u i−1 ),for v L +1≤ i≤v H

wherein {tilde over (d)} i denotes the updated normalized number of codeword values, d i denotes the normalized number of required codewords, if D denotes the sum of d i values, then U=1−D, v H denotes a maximum input pixel value, v L denotes a minimum input pixel value, and

u
i

=

U
·

(

i
-

v
L

v
H

-

v
L

)

α

,

for
⁢

⁢

⁢

v
L

≤
i
≤

v
H

,

Assignees

• Dolby Laboratories Licensing Corp

Inventors

• KHERADMAND Amin
• Su Guan-Ming
• Li Cheng-Chi

Classifications

• G06T5/20

  using local operators · CPC title

• G06T2207/20208

  High dynamic range [HDR] image processing · CPC title

• G06T2207/10024

  Color image · CPC title

• G06T2207/20024

  Filtering details · CPC title

• G06T7/11

  Region-based segmentation · CPC title