# US 2025/0148653 A1 · Compressed media transmission and high quality regeneration

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

- Number: US 2025/0148653 A1
- Document ID: US20250148653A1
- Status: Published application
- Kind: A1
- Application number: 18/500,484
- Filed: 2023-11-02
- Published: 2025-05-08
- Earliest claimed: 2023-11-02
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Sathya Santhar; Sridevi Kannan; Sarbajit K. Rakshit; Samuel Mathew Jawaharlal
- Web page: https://patentgrep.com/patent/US20250148653A1

## CPC

- G06T 9/002: {using neural networks}
- G06T 3/4046: using neural networks
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

An embodiment for transmitting and regenerating compressed media is provided. The embodiment may include receiving an image selected by a sender to be transmitted. The embodiment may also include sending the image to a pre-trained generative adversarial network (GAN) model. The embodiment may further include identifying a first set of one or more pixels in the image assigned to one or more high priority Class of Service (CoS) groups and a second set of one or more pixels in the image assigned to one or more low priority CoS groups. The embodiment may also include compressing the one or more pixels in the second set. The embodiment may further include transmitting the one or more pixels in the first set and the compressed one or more pixels in the second set to a recipient in an order from a highest priority CoS group to a lowest priority CoS group.

## Claims

### Claim 1

A computer-based method of transmitting and regenerating compressed media, the method comprising:
receiving an image selected by a sender to be transmitted to a recipient;
sending the image to a pre-trained generative adversarial network (GAN) model;
identifying a first set of one or more pixels in the image assigned to one or more high priority Class of Service (CoS) groups and a second set of one or more pixels in the image assigned to one or more low priority CoS groups based on a pre-defined grouping of the one or more pixels in the first set and the one or more pixels in the second set;
compressing, by the pre-trained GAN model, the one or more pixels in the second set; and
transmitting the one or more pixels in the first set and the compressed one or more pixels in the second set to the recipient in an order from a highest priority CoS group to a lowest priority CoS group.

### Claim 2 (depends on 1)

The computer-based method of claim 1, further comprising:
regenerating the image at a receiving end of the recipient by reconstructing the compressed one or more pixels in the second set.

### Claim 3 (depends on 1)

The computer-based method of claim 1, further comprising a training phase, wherein the training phase comprises:
receiving an image having a window from a knowledge corpus;
masking a pixel in the window of the image;
sliding a kernel of a pre-defined size over at least a portion of the window including the masked pixel in a forward direction;
generating, by a GAN generator, a predicted pixel for the masked pixel based on one or more surrounding pixels;
determining whether a discriminator of the GAN is able to identify the predicted pixel as real; and
in response to determining the discriminator is able to identify the predicted pixel as real:
generating a reconstruction quality score for the pixel based on a similarity between the pixel and the predicted pixel; and
assigning a CoS priority level to the pixel based on the reconstruction quality score in accordance with a CoS priority table.

### Claim 4 (depends on 3)

The computer-based method of claim 3, wherein the training phase further comprises iterating, for each additional pixel in the window:
masking the additional pixel in the window of the image;
sliding the kernel of the pre-defined size over at least an additional portion of the window including the additional masked pixel in the forward direction;
generating, by the GAN generator, an additional predicted pixel for the additional masked pixel based on one or more additional surrounding pixels;
determining whether the discriminator of the GAN is able to identify the additional predicted pixel as real;
in response to determining the discriminator is able to identify the additional predicted pixel as real:
generating an additional reconstruction quality score for the additional pixel based on a similarity between the additional pixel and the additional predicted pixel; and
assigning an additional CoS priority level to the additional pixel based on the additional reconstruction quality score in accordance with the CoS priority table; and
assigning the pixel and each additional pixel to one or more CoS priority groups based on the CoS priority level and the additional CoS priority level.

### Claim 5 (depends on 1)

The computer-based method of claim 1, wherein the one or more pixels in the first set are transmitted before the compressed one or more pixels in the second set.

### Claim 6 (depends on 1)

The computer-based method of claim 1, wherein the one or more pixels in the first set have a lower reconstruction quality score than the compressed one or more pixels in the second set.

### Claim 7 (depends on 1)

The computer-based method of claim 1, wherein the reconstruction quality score of the compressed one or more pixels in the second set is between 41% and 100%.

### Claim 8

A computer system, the computer system comprising:
one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage medium, and program instructions stored on at least one of the one or more computer-readable tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more computer-readable memories, wherein the computer system is capable of performing a method comprising:
receiving an image selected by a sender to be transmitted to a recipient;
sending the image to a pre-trained generative adversarial network (GAN) model;
identifying a first set of one or more pixels in the image assigned to one or more high priority Class of Service (CoS) groups and a second set of one or more pixels in the image assigned to one or more low priority CoS groups based on a pre-defined grouping of the one or more pixels in the first set and the one or more pixels in the second set;
compressing, by the pre-trained GAN model, the one or more pixels in the second set; and
transmitting the one or more pixels in the first set and the compressed one or more pixels in the second set to the recipient in an order from a highest priority CoS group to a lowest priority CoS group.

### Claim 9 (depends on 8)

The computer system of claim 8, the method further comprising:
regenerating the image at a receiving end of the recipient by reconstructing the compressed one or more pixels in the second set.

### Claim 10 (depends on 8)

The computer system of claim 8, the method further comprising a training phase, wherein the training phase comprises:
receiving an image having a window from a knowledge corpus;
masking a pixel in the window of the image;
sliding a kernel of a pre-defined size over at least a portion of the window including the masked pixel in a forward direction;
generating, by a GAN generator, a predicted pixel for the masked pixel based on one or more surrounding pixels;
determining whether a discriminator of the GAN is able to identify the predicted pixel as real; and
in response to determining the discriminator is able to identify the predicted pixel as real:
generating a reconstruction quality score for the pixel based on a similarity between the pixel and the predicted pixel; and
assigning a CoS priority level to the pixel based on the reconstruction quality score in accordance with a CoS priority table.

### Claim 11 (depends on 10)

The computer system of claim 10, wherein the training phase further comprises iterating, for each additional pixel in the window:
masking the additional pixel in the window of the image;
sliding the kernel of the pre-defined size over at least an additional portion of the window including the additional masked pixel in the forward direction;
generating, by the GAN generator, an additional predicted pixel for the additional masked pixel based on one or more additional surrounding pixels;
determining whether the discriminator of the GAN is able to identify the additional predicted pixel as real;
in response to determining the discriminator is able to identify the additional predicted pixel as real:
generating an additional reconstruction quality score for the additional pixel based on a similarity between the additional pixel and the additional predicted pixel; and
assigning an additional CoS priority level to the additional pixel based on the additional reconstruction quality score in accordance with the CoS priority table; and
assigning the pixel and each additional pixel to one or more CoS priority groups based on the CoS priority level and the additional CoS priority level.

### Claim 12 (depends on 8)

The computer system of claim 8, wherein the one or more pixels in the first set are transmitted before the compressed one or more pixels in the second set.

### Claim 13 (depends on 8)

The computer system of claim 8, wherein the one or more pixels in the first set have a lower reconstruction quality score than the compressed one or more pixels in the second set.

### Claim 14 (depends on 8)

The computer system of claim 8, wherein the reconstruction quality score of the compressed one or more pixels in the second set is between 41% and 100%.

### Claim 15

A computer program product, the computer program product comprising:
one or more computer-readable tangible storage medium and program instructions stored on at least one of the one or more computer-readable tangible storage medium, the program instructions executable by a processor capable of performing a method, the method comprising:
receiving an image selected by a sender to be transmitted to a recipient;
sending the image to a pre-trained generative adversarial network (GAN) model;
identifying a first set of one or more pixels in the image assigned to one or more high priority Class of Service (CoS) groups and a second set of one or more pixels in the image assigned to one or more low priority CoS groups based on a pre-defined grouping of the one or more pixels in the first set and the one or more pixels in the second set;
compressing, by the pre-trained GAN model, the one or more pixels in the second set; and
transmitting the one or more pixels in the first set and the compressed one or more pixels in the second set to the recipient in an order from a highest priority CoS group to a lowest priority CoS group.

### Claim 16 (depends on 15)

The computer program product of claim 15, the method further comprising:
regenerating the image at a receiving end of the recipient by reconstructing the compressed one or more pixels in the second set.

### Claim 17 (depends on 15)

The computer program product of claim 15, the method further comprising a training phase, wherein the training phase comprises:
receiving an image having a window from a knowledge corpus;
masking a pixel in the window of the image;
sliding a kernel of a pre-defined size over at least a portion of the window including the masked pixel in a forward direction;
generating, by a GAN generator, a predicted pixel for the masked pixel based on one or more surrounding pixels;
determining whether a discriminator of the GAN is able to identify the predicted pixel as real; and
in response to determining the discriminator is able to identify the predicted pixel as real:
generating a reconstruction quality score for the pixel based on a similarity between the pixel and the predicted pixel; and
assigning a CoS priority level to the pixel based on the reconstruction quality score in accordance with a CoS priority table.

### Claim 18 (depends on 17)

The computer program product of claim 17, wherein the training phase further comprises iterating, for each additional pixel in the window:
masking the additional pixel in the window of the image;
sliding the kernel of the pre-defined size over at least an additional portion of the window including the additional masked pixel in the forward direction;
generating, by the GAN generator, an additional predicted pixel for the additional masked pixel based on one or more additional surrounding pixels;
determining whether the discriminator of the GAN is able to identify the additional predicted pixel as real;
in response to determining the discriminator is able to identify the additional predicted pixel as real:
generating an additional reconstruction quality score for the additional pixel based on a similarity between the additional pixel and the additional predicted pixel; and
assigning an additional CoS priority level to the additional pixel based on the additional reconstruction quality score in accordance with the CoS priority table; and
assigning the pixel and each additional pixel to one or more CoS priority groups based on the CoS priority level and the additional CoS priority level.

### Claim 19 (depends on 15)

The computer program product of claim 15, wherein the one or more pixels in the first set are transmitted before the compressed one or more pixels in the second set.

### Claim 20 (depends on 15)

The computer program product of claim 15, wherein the one or more pixels in the first set have a lower reconstruction quality score than the compressed one or more pixels in the second set.
