# US 12,217,398 B1 · Apparatus and method for single image dehazing considering haze type classification

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

- Number: US 12,217,398 B1
- Document ID: US12217398B1
- Status: Granted
- Kind: B1
- Application number: 18/658,673
- Filed: 2024-05-08
- Granted: 2025-02-04
- Earliest claimed: 2024-05-08
- Assignees: Imam Mohammad Ibn Saud Islamic University
- Applicants: Imam Mohammad Ibn Saud Islamic University
- Inventors: Md Tanvir Islam; Ik Hyun Lee; Abdul Khader Jilani Saudagar; Abdullah Altameem; Mohammed Abaoud; Khan Muhammad
- Web page: https://patentgrep.com/patent/US12217398B1

## CPC

- G06T 5/60: using machine learning, e.g. neural networks
- G06T 5/73: Deblurring; Sharpening
- 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 10/87: {using selection of the recognition techniques, e.g. of a classifier in a multiple classifier system}
- G06V 10/945: {User interactive design; Environments; Toolboxes}
- G06V 20/50: Context or environment of the image
- G06T 2207/10032: Satellite or aerial image; Remote sensing
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/20092: Interactive image processing based on input by user
- G06T 2207/30181: Earth observation
- G06T 2207/30232: Surveillance
- G06T 2207/30252: Vehicle exterior; Vicinity of vehicle
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A system for intelligent image dehazing based on haze type classification. The system receives a hazy image and classifies the haze type using either a Single Selective Classifier (SSC) or a Hybrid Conditional Classifier (HCC), based on user preference. If SSC is selected, a specific single model is used for classification. If HCC is selected, all pre-trained and pre-defined models are used. The system then selects a suitable specialized dehazer based on the predicted haze type and applies it to the input image to mitigate the effects of the haze. The system is flexible, allowing for adjustments to various parameters as per the users' requirements and target application domains.

## Claims

### Claim 1

A method of dehazing an image having a haze condition and a haze-type class, comprising:
creating a plurality of image sets by splitting a plurality of ground truth images, wherein the plurality of image sets includes an outdoor image set, a street image set, a farmland image set, and a satellite image set;
transitioning each image set of the plurality of image sets by overlaying a synthetic haze condition to obtain a hazy dataset including a plurality of transitioned hazy images having a plurality of haze conditions, wherein the synthetic haze condition includes a haze intensity level;
training a plurality of haze-type classifier models and a plurality of specialized dehazer models with the hazy dataset;
determining a classifier model from the plurality of haze-type classifier models based on a classifier selector, wherein the classifier selector is a hybrid conditional classifier (HCC);
processing the image in accordance with a requirement of each haze-type classifier model of the plurality of haze-type classifier models to obtain a processed image;
determining the haze-type class of the image by the classifier model based on the processed image with the classifier model selected from the plurality of haze-type classifier models;
determining a dehazer from the plurality of specialized dehazer models based on the haze-type class; and
dehazing the image by removing the haze condition with the dehazer.

### Claim 2 (depends on 1)

The method of claim 1, wherein the plurality of haze conditions of the plurality of transitioned hazy images includes a cloud, a fog, an environmental haze, and a rain, wherein the environmental haze includes a carbon emission, a wildfire smoke, and a cloud.

### Claim 3 (depends on 2)

The method of claim 2, wherein the plurality of haze conditions of the plurality of transitioned hazy images in the outdoor image set, the street image set, and the farmland image set further includes a smoke condition.

### Claim 4 (depends on 3)

The method of claim 3, wherein the plurality of specialized dehazer models includes a cloud dehazer, a fog dehazer, an environmental haze dehazer, a smoke dehazer, and a rain dehazer.

### Claim 5 (depends on 1)

The method of claim 1, wherein the determining the classifier model further comprises:
generating an HCC information including a first vector set containing predicted classes, a second vector set containing prediction probabilities, a number of classes (N), and a number of haze-type models (M) for each haze-type classifier model of the plurality of haze-type classifier models with the inference;
passing the HCC information for each haze-type classifier model of the plurality of haze-type classifier models through a condition block to obtain a final predicted class; and
selecting the final predicted class as the classifier model.

### Claim 6 (depends on 5)

The method of claim 5, wherein the condition block comprises a first condition block, a second condition block, a third condition block, a fourth condition block, and a fifth condition block.

### Claim 7 (depends on 6)

The method of claim 6, wherein the first condition block having a first condition of is configured to:
pass the HCC information for each haze-type classifier model of the plurality of haze-type classifier models to the first condition block;
determine whether the first condition is satisfied for each haze-type classifier model of the plurality of haze-type classifier models;
in response to the determining whether the first condition is satisfied, when the first condition is satisfied, pass the HCC information for each haze-type classifier model of the plurality of haze-type classifier models to the third condition block;
in response to the determining whether the first condition is satisfied, when the first condition is not satisfied, pass the HCC information for each haze-type classifier model of the plurality of haze-type classifier models to the second condition block;
wherein the first condition is represented by a first equation of
Count
⁢
C
*
=
N
M
,
wherein the C* represent the haze-type class.

### Claim 8 (depends on 7)

The method of claim 7, wherein the second condition block having a second condition is configured to:
determine whether the second condition is satisfied for each haze-type classifier model of the plurality of haze-type classifier models;
in response to the determining whether the second condition is satisfied, when the second condition is satisfied, pass the HCC information for each haze-type classifier model of the plurality of haze-type classifier models to the third condition block;
in response to the determining whether the second condition is satisfied, when the second condition is not satisfied, pass the HCC information for each haze-type classifier model of the plurality of haze-type classifier models to the fourth condition block;
wherein the second condition is represented by a second equation of
P
i
,
PC
i
=
1
N
,
wherein the Pi, PCi represents the prediction probabilities, wherein the i is a first haze-type classifier model index and PCi is a predicted class of an ith haze-type classifier model.

### Claim 9 (depends on 8)

The method of claim 8, wherein the third condition block is configured to determine the final predicted class based on the second vector of the HCC information.

### Claim 10 (depends on 8)

The method of claim 8, wherein the fourth condition block having a fourth condition is configured to:
determine whether the fourth condition is satisfied for each haze-type classifier model of the plurality of haze-type classifier models;
in response to the determining whether the fourth condition is satisfied, when the fourth condition is satisfied, pass the HCC information for each haze-type classifier model of the plurality of haze-type classifier models to the fifth condition block;
in response to the determining whether the fourth condition is satisfied, when the fourth condition is not satisfied, pass the HCC information for each haze-type classifier model of the plurality of haze-type classifier models to the third condition block;
wherein the fourth condition is represented by a third equation of Pi,j≠Pi,k, wherein the j is a first predicted class index and k is a second predicted class index and wherein j≠k.

### Claim 11 (depends on 10)

The method of claim 10, wherein the fifth condition block is configured to determine the final predicted class based on an accuracy of the plurality of haze-type classifier models.

### Claim 12 (depends on 1)

The method of claim 1, wherein the image is obtained from an autonomous vehicle having a camera or from a surveillance camera.