# US 2025/0095125 A1 · Neural networks for dynamic range conversion and display management of images

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0095125 A1
- Document ID: US20250095125A1
- Status: Published application
- Kind: A1
- Application number: 18/291,578
- Filed: 2022-07-22
- Published: 2025-03-20
- Earliest claimed: 2021-07-29
- Assignees: Dolby Laboratories Licensing Corporation
- Applicants: Dolby Laboratories Licensing Corporation
- Inventors: Robert Wanat; Anustup Kumar Atanu Choudhury; Robin Atkins
- Web page: https://patentgrep.com/patent/US20250095125A1

## CPC

- G06T 5/92: based on global image properties
- G06T 5/60: using machine learning, e.g. neural networks
- G06V 10/60: relating to illumination properties, e.g. using a reflectance or lighting model
- G06V 10/758: {Involving statistics of pixels or of feature values, e.g. histogram matching}
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/20208: High dynamic range [HDR] image processing
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Methods and systems for dynamic range conversion and display mapping of standard dynamic range (SDR) images onto high dynamic range (HDR) displays are described. Given an SDR input image, a processor generates an intensity (luminance) image and optionally a base layer image and a detail layer image. A first neural network uses the intensity image to predict statistics of the SDR image in a higher dynamic range. These predicted statistics together with the original image statistics of the input image are used to derive an optimal tone-mapping curve to map the input SDR image onto an HDR display. Optionally, a second neural network, using the intensity image and the detail layer image, can generate a residual detail layer image in a higher dynamic range to enhance the tone-mapping of the base layer image into the higher dynamic range.

## Claims

### Claim 1

A method for dynamic range conversion and display mapping, the method comprising:
accessing an input image in a first dynamic range and a first spatial resolution;
generating an intensity image based on the input image;
applying the intensity image to a first neural network to generate predicted statistics of the intensity image when mapped in a second dynamic range higher than the first dynamic range; and
generating a tone-mapping curve based on statistics of the intensity image and the predicted statistics; and
applying the tone-mapping curve to the input image to extend the dynamic range of the input image.

### Claim 2 (depends on 1)

The method of claim 1, further comprising generating, based on the input image and the tone-mapping curve, a mapped output image for display on a display with a target dynamic range.

### Claim 3 (depends on 1)

The method of claim 1, further comprising applying the tone-mapping curve to display the input image on a display with a target dynamic range different than the second dynamic range.

### Claim 4 (depends on 1)

The method of claim 1, wherein the statistics of the intensity image comprise intensity values of the intensity image in the first dynamic range, and the predicted statistics comprise predicted intensity values in the second dynamic range.

### Claim 5 (depends on 1)

The method of claim 1, wherein the statistics of the intensity image comprise minimum, average, and maximum intensity values of the intensity image in the first dynamic range, and the predicted statistics comprise predicted minimum, average, and maximum intensity values in the second dynamic range.

### Claim 6 (depends on 1)

The method of claim 1, wherein the first neural network comprises four layers, wherein
the first layer comprises a first filter bank sized 3×3×1×4, with a stride of 2 and 4 biases, followed by a first rectified linear unit (ReLU) activation function;
the second layer comprises a second filter bank sized 3×3×1×8, with a stride of 2 and 8 biases, followed by a second ReLU;
the third layer comprises a third filter bank sized 7×7×2×16, with a stride of 5 and 16 biases, followed by a third ReLU; and
the fourth layer comprises a fourth filter bank sized 48×27×16×3, fully connected, with 3 biases and one 1×3 output representing the predicted statistics of the intensity image when mapped in the second dynamic range.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
generating a base layer image and a detail layer image based on the intensity image; and
applying the tone-mapping curve to the base layer image to generate a tone-mapped base layer image in the second dynamic range.

### Claim 8 (depends on 7)

The method of claim 7, wherein the base layer image represents a spatially-blurred, edge-preserved, version of the intensity image and generating the detail layer image comprises computing:
I
DL
(
x
,
y
)
=
I
⁡
(
x
,
y
)
-
I
BL
(
x
,
y
)
*
dg
,
where, at pixel locations (x,y), IDL denotes the detail layer image, IBL denotes the base layer image, I denotes the intensity image, and dg denotes a detail-gain scaler in [0, 1].

### Claim 9 (depends on 7)

The method of claim 7, further comprising:
adding the detail layer image to the tone-mapped base layer image to generate an output image in the second dynamic range.

### Claim 10 (depends on 7)

The method of claim 7, further comprising:
applying the intensity image and the detail layer image into a second neural network to generate a residual layer image in the second dynamic range;
adding the residual layer image to the detail layer image to generate a second detail layer image; and
adding the second detail layer image to the tone-mapped base layer image to generate an output image in the second dynamic range.

### Claim 11 (depends on 10)

The method of claim 10, wherein the second neural network comprises
an input convolution network, followed by five Residual Networks (ResNet) blocks, followed by a final ReLU and an output convolution network.

### Claim 12 (depends on 11)

The method of claim 11, wherein the input convolution network comprises:
an M×N×2 input;
a filter bank of: 3×3×2×32, stride 1, no biases; and
an M×N×32 output; and
the output convolution network comprises:
an M×N×32 input;
a filter bank of: 3×3×32×1, stride 1, no biases; and
an M×N×1 output, wherein M and N are integers.

### Claim 13 (depends on 11)

The method of claim 11, wherein a Residual Network block comprises:
a first ReLU, followed by a first convolution layer, followed by a second ReLU, followed by a second convolution layer, followed by an adder to add the input to the ResNet network block with the output of the second convolution layer to generate a ResNet output, wherein each of the first and second convolutional layers has a 3×3×32×32 filter bank, with no biases, and a stride of 1.

### Claim 14 (depends on 9)

The method of claim 9, further comprising: applying a display mapping process to map the output image to a display in a target dynamic range different than the second dynamic range.

### Claim 15 (depends on 1)

The method of claim 1, further comprising reducing the first spatial resolution of the intensity image before applying it to the first neural network.

### Claim 16 (depends on 1)

An apparatus comprising a processor and configured to perform claim 1.

### Claim 17 (depends on 1)

A non-transitory computer-readable storage medium having stored thereon computer-executable instruction for executing a method with one or more processors in accordance with claim 1.
