# US 2021/0241024 A1 · Jpeg classifier

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

- Number: US 2021/0241024 A1
- Document ID: US20210241024A1
- Status: Published application
- Kind: A1
- Application number: 17/164,938
- Filed: 2021-02-02
- Published: 2021-08-05
- Earliest claimed: 2020-02-02
- Assignees: DSP Group Ltd.
- Applicants: Emza Visual Sense Ltd.
- Inventors: TAL HENDEL
- Web page: https://patentgrep.com/patent/US20210241024A1

## CPC

- H04N 19/48: using compressed domain processing techniques other than decoding, e.g. modification of transform coefficients, variable length coding [VLC] data or run-length data (motion estimation in a transform domain H04N19/547; processing of decoded motion vectors H04N19/513)
- G06F 18/253: {of extracted features}
- 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 10/764: using classification, e.g. of video objects
- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06V 10/806: {of extracted features}
- G06V 20/80: Recognising image objects characterised by unique random patterns
- G06V 30/19173: {Classification techniques}
- G06V 30/2504: {Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches}
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A method, device and computer program product, the method comprising: obtaining access to a classifier trained upon a multiplicity of sets of decoded coefficients; obtaining a set of block coefficients associated with at least a part of the compressed image; and applying the classifier to the set of block coefficients, to obtain a classification of the compressed image.

## Claims

### Claim 1

A method for classifying a compressed image, comprising:
obtaining access to a classifier trained upon a multiplicity of sets of decoded coefficients;
obtaining a set of block coefficients associated with at least a part of the compressed image; and
applying the classifier to the set of block coefficients, to obtain a classification of the compressed image.

### Claim 2 (depends on 1)

The method of claim 1, wherein the set of block coefficients is received from a capture device configured for:
compressing an image to determine the block coefficients for a captured image; and
outputting at least a part of the block coefficients no later than the image is fully compressed.

### Claim 3 (depends on 1)

The method of claim 1, wherein the set of block coefficients is obtained by:
obtaining a compressed image, comprising decoded block coefficients of an original image; and
partially decompressing the compressed image to obtain the decoded block coefficients as the set of block coefficients.

### Claim 4 (depends on 3)

The method of claim 3, wherein partially decompressing the compressed image and applying the classifier to the block coefficients takes at most 50% of the processing steps required for full decompression and classification of the compressed image.

### Claim 5 (depends on 3)

The method of claim 3, wherein partially decompressing the compressed image and applying the classifier to the block coefficients takes at most 10% of the processing steps required for full decompression and classification of the compressed image.

### Claim 6 (depends on 3)

The method of claim 3, wherein partially decompressing the compressed image and applying the classifier to the quantized block coefficients is performed in hardware.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
obtaining from the classifier a probability of the classification; and
subject to the probability exceeding a predetermined threshold, performing full classification of the compressed image.

### Claim 8 (depends on 7)

The method of claim 7, wherein performing the full classification comprises:
fully decompressing the compressed image to obtain a second set of block coefficients;
de-quantizing the second set of block coefficients, to obtain de-quantized coefficients;
inverse transforming the de-quantized coefficients to obtain pixel values of the original image;
extracting feature vectors form the original image; and
classifying the original image using the feature vectors.

### Claim 9 (depends on 1)

The method of claim 1, wherein applying the classifier to the set of block coefficients comprises:
combining the set of block coefficients in a first predetermined block combination to obtain first combined block coefficients, and wherein the classifier is applied to the first combined block coefficients; and further comprising:
obtaining a first classification probability from applying the classifier on the combined block coefficients;
combining the set of block coefficients in a second predetermined block combination to obtain second combined block coefficients;
applying the classifier to the second combined block coefficients, to obtain a classification of the compressed image and a second probability; and
performing the full classification of the compressed image, subject to the maximum between the first probability and the second probability exceeding the predetermined threshold.

### Claim 10 (depends on 9)

The method of claim 9, wherein the first predetermined block combination and the second predetermined block combination are in accordance with expected characteristics of the original image.

### Claim 11 (depends on 9)

The method of claim 9, wherein performing the full classification comprises:
fully decompressing the compressed image to obtain a second set of block coefficients;
de-quantizing the second set of block coefficients, to obtain de-quantized coefficients;
inverse transforming the de-quantized coefficients to obtain pixel values of the original image;
extracting feature vectors from the original image; and
classifying the original image using the feature vectors.

### Claim 12 (depends on 1)

The method of claim 1, wherein the method is used for security or privacy protection purposes.

### Claim 13 (depends on 1)

The method of claim 1, wherein the compressed image is a Joint Photographic Experts Group (JPEG) image.

### Claim 14

A device comprising:
an image capture device for capturing a captured image;
a memory device for storing a compressed image being the captured image as compressed, the compressed image comprising a set of block coefficients; and
a processor, executing:
obtaining access to a classifier trained upon a multiplicity of sets of decoded coefficients;
obtaining a set of block coefficients associated with at least a part of the compressed image; and
applying the classifier to the set of block coefficients, to obtain a classification of the compressed image.

### Claim 15 (depends on 14)

The device of claim 14, wherein the set of block coefficients is received from a capture device configured for:
compressing an image to determine the block coefficients for a captured image; and
outputting at least a part of the block coefficients no later than the image is compressed.

### Claim 16 (depends on 14)

The device of claim 14, wherein the set of block coefficients is obtained by:
obtaining a compressed image, comprising decoded block coefficients of an original image; and
partially decompressing the compressed image to obtain the decoded block coefficients as the set of block coefficients.

### Claim 17 (depends on 14)

The device of claim 14, wherein partially decompressing the compressed image and applying the classifier to the block coefficients takes at most 10% of the processing steps required for full decompression and classification of the compressed image.

### Claim 18 (depends on 14)

The device of claim 14, wherein applying the classifier to the set of block coefficients comprises:
combining the set of block coefficients in a first predetermined block combination to obtain first combined block coefficients, and wherein the classifier is applied to the first combined block coefficients; and further comprising:
obtaining a first classification probability from applying the classifier on the combined block coefficients;
combining the set of block coefficients in a second predetermined block combination to obtain second combined block coefficients;
applying the classifier to the second combined block coefficients, to obtain a classification of the compressed image and a second probability; and
performing the full classification of the compressed image, subject to the maximum between the first probability and the second probability exceeding the predetermined threshold.

### Claim 19 (depends on 14)

The device of claim 14, wherein the compressed image is a Joint Photographic Experts Group (JPEG) image.

### Claim 20

A computer program product comprising a non-transitory computer readable storage medium retaining program instructions, which program instructions when read by a processor, cause the processor to perform a method comprising:
obtaining access to a classifier trained upon a multiplicity of sets of decoded coefficients;
obtaining a set of block coefficients associated with at least a part of the compressed image; and
applying the classifier to the set of block coefficients, to obtain a classification of the compressed image.