# US 2018/0005078 A1 · Configurable histogram-of-oriented gradients (hog) processor

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

- Number: US 2018/0005078 A1
- Document ID: US20180005078A1
- Status: Published application
- Kind: A1
- Application number: 15/198,360
- Filed: 2016-06-30
- Published: 2018-01-04
- Earliest claimed: 2016-06-30
- Assignees: Apple Inc.
- Applicants: Apple Inc.
- Inventors: Suk Hwan Lim; Jaewon Shin
- Web page: https://patentgrep.com/patent/US20180005078A1

## CPC

- G06V 10/50: by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Embodiments relate to a histogram-of-oriented gradients (HOG) module. The HOG module is implemented in hardware rather than software. The HOG module applies an algorithm to an image to identify gradient orientation in localized portions of the image. The HOG module creates a histogram-of orientation gradients based on the identified gradient orientations.

## Claims

### Claim 1

A configurable histogram-of-oriented gradients (HOG) module comprising:
a plurality of registers configured to store configuration values including a first configuration value;
a HOG engine circuit coupled to the plurality of registers to receive the first configuration value, the HOG engine circuit configured to receive image data and generate histograms of different types of HOG features of the image data based on at least the first configuration value received from the plurality of registers, the different type of HOG features of the image data including a total number of edges in the image data with a magnitude less than a threshold; and
an output interface circuit coupled to the HOG engine circuit, the output interface circuit configured to output the histograms of different types of HOG features generated by the HOG engine circuit to a computer processing unit coupled to the HOG module.

### Claim 2 (depends on 1)

The configurable HOG module of claim 1, further comprising:
an input interface circuit configured to receive the image data and provide the image data to the HOG engine circuit.

### Claim 3 (depends on 1)

The configurable HOG module of claim 1, wherein the HOG engine circuit comprises:
a normalizer configured to normalize at least one of the different types of HOG features according to a second configuration value stored in the plurality of registers.

### Claim 4 (depends on 3)

The configurable HOG module of claim 3, wherein the different types of HOG features of the image data include non-normalized HOG features and the HOG engine circuit generates non-normalized HOG features of the image data based on a third configuration value stored in the plurality of registers.

### Claim 5 (depends on 3)

The configurable HOG module of claim 3, wherein the different types of HOG features of the image data include a summation of magnitudes of edges included in the image data and the HOG engine circuit is configured to calculate the summation by (i) calculating a magnitude of each edge included in the image data and (ii) summing the magnitudes of the edges based on a third configuration value stored in the plurality of registers.

### Claim 6 (depends on 5)

The configurable HOG module claim 5, wherein the summation is based on normalized HOG features of the image data.

### Claim 7 (depends on 5)

The configurable HOG module claim 5, wherein the summation is based on non-normalized HOG features of the image data.

### Claim 8

(canceled)

### Claim 9 (depends on 1)

The configurable HOG module of claim 1, wherein the HOG engine circuit is configured to determine the total number of edges by:
calculating a magnitude of each edge in the image data; and
determining the total number of edges having magnitudes less the threshold.

### Claim 10 (depends on 1)

The configurable HOG module of claim 1, further comprising a patch mode processor coupled to the HOG engine circuit, the patch mode processor configured to generate a histogram of gradient features for each of a plurality of patches included in the image data.

### Claim 11 (depends on 10)

The configurable HOG module of claim 10, wherein each of the plurality of image patches is extracted from different source images.

### Claim 12 (depends on 10)

The configurable HOG module of claim 10, wherein each of the plurality of image patches is extracted from a single source image.

### Claim 13

A method, comprising:
setting first operation parameters of components of a histogram-of-oriented gradients (HOG) module based on a first configuration value stored in a plurality of configuration registers included in the HOG module;
receiving image data for processing by the HOG module;
generating, by the HOG module, histograms of different types of HOG features of the received image data based on at least the first operation parameters set according to the first configuration value stored in the plurality of configuration registers, the different type of HOG features of the received image data including a total number of edges in the received image data with a magnitude less than a threshold; and
outputting the generated histograms of different types of HOG features to a computer processing unit coupled to the HOG module.

### Claim 14 (depends on 13)

The method of claim 13, further comprising:
receiving the first configuration value of the plurality of configuration registers from the computer processing unit coupled to the HOG module.

### Claim 15 (depends on 13)

The method of claim 13, further comprising:
normalizing at least one of the different types of HOG features according to a second configuration value stored in the plurality of configuration registers.

### Claim 16 (depends on 15)

The method of claim 15, wherein the different types of HOG features of the image data include non-normalized HOG features and the method further comprising:
generating non-normalized HOG features of the image data based on a third configuration value stored in the plurality of registers.

### Claim 17 (depends on 15)

The method of claim 15, wherein the different types of HOG features of the image data include a summation of magnitudes of edges included in the image data and the method further comprising:
calculating a magnitude of each edge included in the image data; and
summing the magnitudes of the edges based on a third configuration value stored in the plurality of configuration registers.

### Claim 18 (depends on 17)

The method of claim 17, wherein the summation is based on either normalized HOG features of the image data or non-normalized HOG features of the image data.

### Claim 19 (depends on 13)

The method of claim 13 further comprising:
calculating a magnitude of each edge in the image data; and
determining the total number of edges having magnitudes less the threshold.

### Claim 20

A method for processing an image, comprising:
enabling a patch mode of a histogram-of-oriented gradients (HOG) module based on a configuration value stored in a plurality of configuration registers included in the HOG module;
receiving image data that comprises a plurality of image patches responsive to the patch mode;
identifying boundaries of the plurality of image patches;
generating a histogram of gradient features for each of the plurality of image patches based on the identified boundaries; and
outputting the generated histogram of gradient features for each of the plurality of image patches.
