# US 2019/0244330 A1 · Dynamically determining filtering strength for noise filtering in image processing

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

- Number: US 2019/0244330 A1
- Document ID: US20190244330A1
- Status: Published application
- Kind: A1
- Application number: 16/389,785
- Filed: 2019-04-19
- Published: 2019-08-08
- Earliest claimed: 2015-08-31
- Assignees: Apple Inc.
- Applicants: Apple Inc.
- Inventors: Suk Hwan Lim; D. Amnon Silverstein
- Web page: https://patentgrep.com/patent/US20190244330A1

## CPC

- G06T 5/70: Denoising; Smoothing
- G06T 1/20: Processor architectures; Processor configuration, e.g. pipelining
- G06T 7/41: based on statistical description of texture
- H04N 23/81: for suppressing or minimising disturbance in the image signal generation
- H04N 23/843: {Demosaicing, e.g. interpolating colour pixel values}
- G06T 2207/10024: Color image
- G06T 2207/20012: Locally adaptive
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

An image processing pipeline may dynamically determine filtering strengths for noise filtering of image data. Statistics may be collected for an image at an image processing pipeline. The statistics may be accessed and evaluated to generate a filter strength model that maps respective filtering strengths to different portions of the image. A noise filter may determine a filtering strength for image data received at the noise filter according to the filter strength model. The noise filter may then apply a filtering technique according to the determined filtering strength.

## Claims

### Claim at position 1

-20. (canceled)

### Claim 21

An apparatus, comprising:
a memory; and
an image signal processor comprising a statistics collection unit and a noise filter, the image signal processor configured to:
collect, by the statistics collection unit, a plurality of statistics for an image comprising pixel values for a plurality of locations, wherein the pixel values indicate respective colors of pixels at the plurality of locations;
store a filter strength model to the memory that maps respective filtering strengths to different portions of the image based on the plurality of collected statistics;
receive, at the noise filter, image data for the image; and
filter the image data to generate filtered image data according to respective filter strengths for different portions of the image data determined according to the filter strength model accessed from the memory.

### Claim 22 (depends on 21)

The apparatus of claim 21, wherein:
the statistics collection unit is configured to store the plurality of statistics to the memory; and
the apparatus further comprises a controller for the image signal processor, the controller configured to:
access the plurality of statistics in the memory;
generate the filter strength model; and
store the filter strength model to the memory.

### Claim 23 (depends on 22)

The apparatus of claim 22, wherein to generate the filter strength model, the controller is configured to calculate a standard deviation gain map for the image.

### Claim 24 (depends on 23)

The apparatus of claim 23:
wherein the image data comprises a pixel in the image;
wherein to determine a filter strength of the respective filter strengths, the image signal processor is configured to interpolate a gain value according to a location of the pixel in the image corresponding to a location in the standard deviation gain map; and
wherein to filter the image data to generate filtered image data, the image signal processor is configured to apply the gain value to a noise standard deviation value that is utilized to filter the pixel.

### Claim 25 (depends on 21)

The apparatus of claim 21:
wherein the image data is received in a full color encoding that comprises a plurality of color channels; and
wherein to filter the image data to generate filtered image data the image signal processor is configured to apply a filtering technique to one of the color channels and apply a different filtering technique to another one of the color channels.

### Claim 26 (depends on 25)

The apparatus of claim 25, wherein the full color encoding is a YCbCr encoding.

### Claim 27 (depends on 21)

The apparatus of claim 21, wherein the apparatus comprises a mobile computing device.

### Claim 28

A method, comprising:
collecting, by an image signal processor, a plurality of statistics for an image comprising pixel values for a plurality of locations, wherein the pixel values indicate respective colors of pixels at the plurality of locations;
storing to a memory, by the image signal processor, a filter strength model that maps respective filtering strengths to different portions of the image based on the plurality of collected statistics;
receiving, at a noise filter for the image signal processor, image data for the image; and
filtering, at the noise filter, the image data to generate filtered image data according to respective filter strengths for different portions of the image data determined according to the filter strength model accessed from the memory.

### Claim 29 (depends on 28)

The method of claim 28, further comprising collecting, at the image signal processor, a plurality of noise statistics as the plurality of statistics.

### Claim 30 (depends on 29)

The method of claim 29, wherein collecting the plurality of noise statistics comprises:
for one or more non-overlapping blocks of pixels in the image:
calculating spatial gradients for the one or more non-overlapping blocks;
based on the spatial gradients, determining:
dominant gradient orientations for the one or more non-overlapping blocks;
peaks of gradient magnitude for the one or more non-overlapping blocks; and
sums of gradient magnitude for the one or more non-overlapping blocks; and
calculating sums of pixel intensities for the one or more non-overlapping blocks.

### Claim 31 (depends on 30)

The method of claim 30, wherein generating the filter strength model comprises generating a standard deviation gain map based on the dominant gradient orientations, the peaks of gradient magnitude, the sums of gradient magnitude, and the sums of pixel intensities for the one or more non-overlapping blocks.

### Claim 32 (depends on 28)

The method of claim 28, wherein the plurality of statistics comprise color statistics, and wherein generating the filter strength model comprises mapping the respective filtering strengths according to one or more hues present in the image.

### Claim 33 (depends on 28)

The method of claim 28, wherein the filter strength model is a two-dimensional lookup table, and wherein determining the filter strength for the image data comprises:
identifying a corresponding location of the image data in the two-dimensional lookup table; and
interpolating the filter strength with respect to the corresponding location in the two-dimensional lookup table.

### Claim 34 (depends on 28)

The method of claim 28, wherein the image signal processor is implemented as part of a mobile computing device.

### Claim 35

A system, comprising:
a device configured to perform image processing, the device comprising:
an image processing pipeline configured to filter image data of an image received at a noise filter, wherein the image data comprises pixel values for a plurality of locations, and wherein the pixel values indicate respective colors of the pixels at the plurality of locations; and
a controller for the image processing pipeline configured to store, to a memory, a filter strength model that maps respective filtering strengths to different portions of the image based on a plurality of statistics collected for the image by the image processing pipeline;
wherein the noise filter is configured to filter the image data according to respective filter strengths for different portions of the image data determined according to the filter strength model accessed from the memory.

### Claim 36 (depends on 35)

The system of claim 35, wherein the plurality of statistics comprise dominant gradient orientations, peaks of gradient magnitude, sums of gradient magnitude, or sums of pixel intensities for one or more non-overlapping blocks.

### Claim 37 (depends on 36)

The system of claim 36, wherein to store the filter strength model, the controller is configured to generate the filter strength model using a standard deviation gain map.

### Claim 38 (depends on 36)

The system of claim 36, wherein at least one of the respective filter strengths is a gain value, and wherein to filter the image data the image processing pipeline is configured to apply the gain value to a noise standard deviation utilized to filter the image data.

### Claim 39 (depends on 35)

The system of claim 35, wherein, to filter the image data, the noise filter is configured to apply a filtering technique selected from a plurality of filtering techniques that the image processing pipeline is capable of applying to the image data.

### Claim 40 (depends on 35)

The system of claim 35, wherein the system is a mobile computing device.
