1. A method for underwater image classification, restoration, and enhancement, the method comprising the steps of: - acquiring underwater images (101) from one or more imaging sources, including underwater cameras, remotely operated vehicles (ROVs), autonomous underwater vehicles (AUVs), or sonar imaging systems; - preprocessing the acquired underwater images (101) using enhancement techniques, including noise reduction, resizing, normalization, histogram equalization, and contrast-limited adaptive histogram equalization (CLAHE); - training a plurality of deep learning models using annotated underwater images (101) with the application of one or more deep learning techniques, including Convolutional Neural Networks (CNN); - validating and testing the trained deep learning models with a separate dataset of underwater images (101) to evaluate performance metrics, including accuracy, precision, recall, and F1-score; - deploying the validated deep learning models into a practical underwater imaging system for real-time application; and - utilizing the deployed deep learning models to classify objects, enhance image visibility, and provide actionable insights from the underwater images (101).

2. The method as claimed in claim 1, wherein the long short-term memory (LSTM) is trained on paired and unpaired underwater images (101) to learn the transformations required for noise reduction and colour correction, thereby enabling the image enhancement module (102) and Object classification & restoration module (103) to adapt to various underwater conditions, including different lighting and turbidity levels.

3. The method as claimed in claim 1, wherein the training of the deep learning models includes data augmentation techniques to improve model generalizability and robustness.

4. The method as claimed in claim 1, wherein the deployment integrates the deep learning models into cloud-based, edge, or standalone systems for real-time underwater imaging applications.

5. The method as claimed in claim 1, wherein the actionable insights generated include object classification (104), habitat monitoring, infrastructure inspection, and navigation support.

6. A system (100) for underwater image classification, restoration, and enhancement, the system comprising: - an acquisition unit (100a) configured to acquire underwater images (101) from one or more imaging sources, including underwater cameras, ROVs, AUVs, or sonar imaging systems; - a pre-processing unit (100b) configured to preprocess the underwater images (101) using enhancement techniques, including noise reduction, resizing, normalization, histogram equalization, and CLAHE; - a training unit (100c) configured to train a plurality of deep learning models using annotated underwater images (101) with one or more deep learning techniques, including CNNs; - a validation and testing unit (100d) configured to validate and test the trained deep learning models with performance metrics, including accuracy, precision, recall, and F1-score; - a deployment unit (100e) configured to deploy the validated deep learning models into real-world underwater imaging applications; and - an analysis unit (100f) configured to classify objects, enhance image visibility, and generate actionable insights from the underwater images (101).

7. The system (100) as claimed in claim 6, wherein the pre-processing unit (100b) applies colour correction algorithms to mitigate underwater light distortion.

8. The system (100) as claimed in claim 6, wherein the training unit (100c) employs data augmentation techniques to enhance the robustness of the deep learning models.

9. The system (100) as claimed in claim 6, wherein the deployment unit (100e) integrates the deep learning models into a cloud-based, edge, or standalone system for real-time underwater imaging processing.

10. The system (100) as claimed in claim 6, wherein the analysis unit (100f) provides actionable insights, including object classification, habitat monitoring, infrastructure inspection, and navigation support, based on the processed underwater images (101).

Description:FIELD OF THE INVENTION
[001] The present invention is related to image processing, specifically to the enhancement and restoration of underwater images using deep learning models. More particularly, the invention involves the use of advanced deep learning techniques, including CNNs and LSTMs, to improve the visibility and classification accuracy of underwater images in real-time environments.

BACKGROUND OF THE INVENTION
[002] Underwater imaging faces significant challenges due to the optical properties of water, which degrade image quality. Water absorbs and scatters light, resulting in colour distortion where red and yellow wavelengths diminish with depth, leaving blue and green hues dominant. Suspended particles scatter light unpredictably, leading to a loss of contrast and increased visual noise. These issues produce blurry, low-contrast images that are difficult to interpret. The variability of underwater environments—ranging from clear tropical waters to sediment-laden regions—further complicates imaging, as solutions effective in one condition often fail in another. These challenges impact critical applications such as object recognition, marine life classification, underwater archaeology, and pipeline inspection.

[003] Conventionally, underwater image enhancement relied on manual preprocessing and heuristic-based techniques. Histogram equalization, for example, was commonly used to improve contrast, while white balancing corrected dominant colour hues. Denoising filters, such as Gaussian and median filters, reduced noise at the cost of fine detail. Localized methods like Contrast Limited Adaptive Histogram Equalization (CLAHE) improved contrast without over-amplifying noise, but these heuristic methods were limited in their adaptability. An example is US11763426B2, which introduces an underwater image enhancement system designed to address issues such as low contrast and colour distortion. However, like many heuristic-based approaches, this system lacks the adaptability to handle dynamic and highly variable underwater conditions, making it less effective in diverse environments.

[004] The advent of deep learning introduced adaptive and automated solutions to underwater imaging challenges. Deep learning-based enhancement models, such as those leveraging Generative Adversarial Networks (GANs), excel in correcting color and restoring visual clarity. For instance, WaterGAN simulates underwater conditions to train models capable of robust image enhancement across varying environments.

[005] Encoder-decoder architectures like UNet have been widely adopted for noise reduction and contrast enhancement. Classification models, such as convolutional neural networks (CNNs) like ResNet and DenseNet, are widely used for underwater object detection and marine species classification. For example, CN112288658A describes a method based on multi-residual joint learning to enhance underwater images by addressing degradation factors. While this approach improves adaptability and image quality, it relies heavily on the availability of labeled training data and computational resources, making real-time applications challenging.

[006] Efforts to integrate enhancement and classification into unified frameworks are gaining traction, as treating these tasks separately often leads to inefficiencies. Unified systems streamline these processes, improving accuracy and efficiency in real-time applications such as marine exploration and autonomous underwater vehicle (AUV) navigation. For example, US20210217144A1 focuses on underwater image color correction using machine learning to identify and reconstruct colors distorted by water’s optical properties. While this system effectively addresses color correction, it does not fully integrate enhancement and classification processes, limiting its use in comprehensive real-time systems for dynamic underwater environments.

[007] The limitations of the above-mentioned existing designs highlight the need for more innovative methods that can holistically address enhancement, restoration, and classification while being efficient and adaptable across diverse underwater conditions.

OBJECTS OF THE PRESENT INVENTION
[008] The main objective of the present invention is to provide a unified system that integrates image enhancement and object classification & restoration into a single, streamlined framework, eliminating the need for separate pre-processing and classification tools, thereby enhancing efficiency in underwater image analysis.

[009] Another objective of the present invention is to incorporate a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) deep learning model for image enhancement, combining the strengths of CNNs for spatial feature extraction and LSTMs for capturing contextual relationships, ensuring consistent and high-quality image restoration across diverse underwater conditions.

[010] Another objective of the present invention is to make the system adaptable to a wide range of underwater environments by training the LSTM model on both paired and unpaired underwater images, enabling the system to handle varying levels of turbidity, lighting, and depths, and produce consistent enhancement results.

[011] Another objective of the present invention is to enhance object detection capabilities through multi-scale feature extraction, enabling the classification of objects at varying sizes and distances, from small marine species to large underwater structures.

[012] Yet another objective of the present invention is to ensure real-time performance for the classification and restoration module, allowing the system to process and classify underwater images instantly, making it suitable for applications in underwater drones, ROVs, and AUVs that require quick decision-making and immediate feedback.

[013] Yet another objective of the present invention is to offer a system that operates on standard computational hardware with GPU acceleration, ensuring efficient real-time image processing and classification for underwater missions.

[014] Yet another objective of the present invention is to ensure robustness to environmental variations, enabling the system to perform effectively across a range of underwater conditions, from clear to murky waters, providing consistent and reliable results.

SUMMARY OF THE PRESENT INVENTION
[015] The present invention relates to a system and method for underwater image classification, restoration, and enhancement using deep learning models, designed to address critical challenges in underwater imaging, such as visibility degradation, color distortion, and noise caused by light scattering and absorption in water. By leveraging advanced deep learning techniques, this invention enhances image quality and enables accurate object classification and restoration in real time. It integrates two core modules: an Image Enhancement Module and an object classification & restoration module, forming a unified system that ensures seamless and efficient underwater image analysis.

[016] In the preferred embodiment, the image enhancement module improves the visual quality of underwater images by correcting color balance, enhancing contrast, and reducing noise. The image enhancement module combines the spatial feature extraction capabilities of Convolutional Neural Networks (CNNs) with the contextual understanding of Long Short-Term Memory networks (LSTMs). The above-mentioned approach allows the system to produce consistent and high-quality enhancements across varying underwater conditions, including different levels of turbidity, lighting, and depths. The module’s adaptability is ensured by training it on large datasets of paired and unpaired underwater images, resulting in enhanced images with restored colour fidelity and reduced visual noise.

[017] The object classification & restoration module identifies, classifies, and restores underwater objects such as marine life, artifacts, and geological formations. It utilizes a unique CNN-LSTM architecture, with CNNs extracting spatial features and LSTMs capturing sequential and contextual relationships to improve classification accuracy and image restoration. Its multi-scale object detection capability allows for identifying objects of various sizes and distances, ranging from small marine species to large underwater structures.

[018] One of the key advantages of the system is its real-time operation, enabling immediate image enhancement and object classification. This makes it ideal for dynamic underwater applications, such as autonomous underwater vehicles (AUVs), remotely operated vehicles (ROVs), and underwater drones, where quick decision-making is critical. The system is designed to function efficiently on standard computational hardware with GPU acceleration, ensuring accessibility and ease of deployment.

[019] Overall, the invention is highly versatile and adaptable, with applications in marine biology, underwater exploration, and environmental monitoring. It supports tasks such as detecting and tracking marine species, observing environmental changes like coral bleaching or pollution, and identifying shipwrecks and underwater artifacts. By overcoming the limitations of traditional methods, such as their inability to adapt to varying underwater conditions or combine enhancement and classification tasks effectively, this invention offers a cutting-edge, unified solution for underwater imaging needs across scientific and industrial domains.

BRIEF DESCRIPTION OF FIGURES
This invention is illustrated in the accompanying drawings, throughout which like reference letters / numerals indicate corresponding parts in the various figures. The embodiments herein and advantages thereof will be better understood from the following description when read with reference to the following drawings, wherein

[020] FIG. 1 is a block diagram of the system for underwater image classification, restoration and enhancement using deep learning models.
[021] FIG. 2 is a flowchart illustrating the method for underwater image classification, restoration and enhancement using deep learning models.

DETAILED DESCRIPTION OF THE DISCLOSURE
[022] Some embodiments of the present disclosure, illustrating all its features, will now be discussed in detail. It must also be noted that as used herein and in the appended claims, the singular forms "a", "an" and "the" include plural references unless the context clearly dictates otherwise.

[023] Various modifications to the embodiment will be readily apparent to those skilled in the art and the generic principles herein may be applied to other embodiments. However, one of ordinary skill in the art will readily recognize that the present disclosure including the definitions listed here below are not intended to be limited to the embodiments illustrated but is to be accorded the widest scope consistent with the principles and features described herein.

[024] A person of ordinary skill in the art will readily ascertain that the illustrated steps detailed in the figures and here below are set out to explain the exemplary embodiments shown, and it should be anticipated that ongoing technological development will change the manner in which particular functions are performed. These examples are presented herein for purposes of illustration, and not limitation. Further, the boundaries of the functional building blocks have been arbitrarily defined herein for the convenience of the description. Alternative boundaries can be defined so long as the specified functions and relationships thereof are appropriately performed. Alternatives (including equivalents, extensions, variations, deviations, etc., of those described herein) will be apparent to persons skilled in the relevant art(s) based on the teachings contained herein. Such alternatives fall within the scope and spirit of the disclosed embodiments.

[025] Referring to FIG. 1, FIG. 1 illustrates a block diagram of an underwater image classification, restoration, and enhancement system (100), in accordance with an embodiment of the present subject matter. The underwater image classification, restoration, and enhancement system (100) is configured to enhance underwater images (101), classify objects within the underwater images (101), and restore degraded components of the underwater images (101) using deep learning models. The underwater image classification, restoration, and enhancement system (100) comprises underwater images (101), an image enhancement module (102), an object classification and restoration module (103), and enhanced images with classified objects (104).

[026] In an embodiment, the underwater images (101) represent input images captured in underwater environments. The underwater images (101) often suffer from poor visibility, low contrast, color distortion, and haze due to light absorption and scattering in water. The underwater images (101) may include various types of visual data collected through underwater cameras, remotely operated vehicles (ROVs), or other underwater imaging devices.

[027] In an embodiment, the image enhancement module (102) is responsible for processing the underwater images (101) to improve their clarity, brightness, and color balance. The image enhancement module (102) employs techniques such as white balancing, histogram equalization, and dehazing algorithms to address the visual challenges associated with underwater imaging. The image enhancement module (102) ensures that the quality of the underwater images (101) is sufficient for subsequent classification and restoration tasks.

[028] In an embodiment, the object classification and restoration module (103) process the output from the image enhancement module (102) to classify objects within the enhanced underwater images and restore any missing or degraded components of the enhanced underwater images (101). The object classification and restoration module (103) utilize hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) deep learning models to identify and label marine objects such as fish, corals, underwater structures, or debris. Additionally, the object classification and restoration module (103) apply advanced restoration techniques to further enhance the visual quality of the underwater images (101), removing any residual noise or distortions.

[029] In an embodiment, the enhanced images with classified objects (104) represent the final output of the underwater image classification, restoration, and enhancement system (100). The enhanced images with classified objects (104) are visually improved versions of the underwater images (101) that also include labeled objects for applications such as marine exploration, biodiversity studies, and environmental monitoring.

[030] The underwater image classification, restoration, and enhancement system (100) utilizes advanced deep learning methodologies to address the specific challenges of underwater imaging. It is to be noted by a person skilled in the art that the specific configurations and algorithms implemented in the underwater image classification, restoration, and enhancement system (100) are illustrative and not restrictive to the scope of the present claimed subject matter.

[031] In one of the embodiments the underwater image classification, restoration, and enhancement system (100) for underwater image classification, restoration, and enhancement, the system comprising, an acquisition Unit (100a), a preprocessing unit (100b), a training unit (100c), a validation and testing Unit (100d), a deployment Unit (100e) and an analysis Unit (100f). The said acquisition unit (100a) configured to acquire underwater images (101) from one or more imaging sources, including underwater cameras, ROVs, AUVs, or sonar imaging systems. The preprocessing unit (100b) configured to preprocess the underwater images (101) using enhancement techniques, including noise reduction, resizing, normalization, histogram equalization, and Contrast Limited Adaptive Histogram Equalization (CLAHE). The training unit (100c) configured to train a plurality of deep learning models using annotated underwater images (101) with one or more deep learning techniques, including Convolutional Neural Network (CNNs). The validation and testing unit (100d) configured to validate and test the trained deep learning models with performance metrics, including accuracy, precision, recall, and F1-score. The deployment unit (100e) configured to deploy the validated deep learning models into real-world underwater imaging applications. The analysis unit (100f) configured to classify objects, enhance image visibility, and generate actionable insights from the underwater images (101).

[032] Referring to FIG. 2, FIG. 2 illustrates a workflow chart (200) that details the step-by-step process of underwater image classification, restoration, and enhancement system (100) using deep learning models, in accordance with an embodiment of the present subject matter. The workflow chart (200) outlines the comprehensive methodology adopted by the underwater image classification, restoration, and enhancement system (100) to process underwater images (101) and achieve enhanced outputs with classified objects (104).

[033] The workflow begins with step (201), which represents the input from the Underwater (UW) Image Dataset. The said underwater (UW) image dataset consists of underwater images (101) captured in hazy conditions and stored as Red-Green-Blue (RGB) images, as represented by step (202). The said underwater images (101) often suffer from poor visibility, low contrast, and color distortion due to light absorption and scattering in the underwater environment.

[034] In step (203), pre-processing is performed on the hazy Red-Green-Blue (RGB) images to remove noise artifacts and improve the overall quality of the underwater images (101). Image denoising in the step (203) ensures the suitability of the underwater images (101) for subsequent enhancement processes. The enhancement processes is followed by step (204), where the underwater images (101) are resized to standardized dimensions to ensure compatibility with the hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model used in later stages.

[035] The workflow chart (200) then proceeds to step (205), where two key techniques, White Balance and Contrast Limited Adaptive Histogram Equalization (CLAHE), are applied to the underwater images (101). White Balance corrects color distortions by compensating for the absorption of red wavelengths underwater, while Contrast Limited Adaptive Histogram Equalization (CLAHE) enhances the local contrast of the underwater images (101), ensuring improved visibility of finer details. Step (206) employs the dark channel prior algorithm, which effectively reduces haze in the underwater images (101) by leveraging statistical properties of haze-free images, thereby improving underwatwer image (101) clarity.

[036] In step (207), Non-Anomalous Information (NAI) identification is performed. The said step (207) ensures the identification of key features or regions within the enhanced images that are relevant for classification and restoration tasks. The workflow chart (200) then splits the processed images into two subsets: training samples, represented by step (208), and testing samples, represented by step (209). Training samples are used to optimize the parameters of the hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model, as part of the training process detailed in step (210) and step (211). The hybrid model combines the spatial feature extraction capabilities of the Convolutional Neural Network with the temporal sequence learning abilities of the Long Short-Term Memory (LSTM).

[037] The step (212) involves testing the trained hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model using the testing samples. The said step (212) evaluates the hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM)'s performance on unseen data to ensure reliability and accuracy in real-world applications. Following the testing phase, step (213) represents the generation of enhanced underwater images with classified objects (104). The said step (213) produces high-quality outputs suitable for applications such as marine exploration, biodiversity studies, and environmental monitoring.

[038] The final stages of the workflow involve performance analysis, represented by step (214), and the generation of results, represented by step (215). Performance analysis evaluates the overall effectiveness of the underwater image classification, restoration, and enhancement system (100) using metrics such as accuracy, Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index Measure (SSIM). The results are compiled into a detailed report, providing a comprehensive assessment of the underwater image classification, restoration, and enhancement system (100).

[039] The workflow chart (200) demonstrates the seamless integration of preprocessing techniques, enhancement algorithms, and deep learning methodologies to address the specific challenges associated with underwater imaging. It is to be noted by a person skilled in the art that the description of the workflow chart (200) provided herein is illustrative and does not limit the underwater image classification, restoration, and enhancement system (100) to a specific implementation or sequence of steps.
[040] As far as the working of the underwater image classification, restoration, and enhancement system (100) (refer to FIG. 1 and FIG. 2) of the present invention is concerned, the system operates seamlessly by integrating its various components for enhancing underwater images, classifying objects within them, and restoring degraded visual components.

[041] As per the illustrated and preferred embodiment, the underwater image classification, restoration, and enhancement system (100) initiates its operation by receiving input images from the Underwater (UW) Image Dataset, as represented by step (201). The said underwater images (101) are typically captured using underwater imaging devices such as remotely operated vehicles (ROVs) or underwater cameras and often suffer from poor visibility, color distortion, and haze due to light scattering and absorption underwater. The received underwater images, stored as hazy Red-Green-Blue (RGB) images, are represented by step (202).

[042] The underwater images (101) undergo preprocessing in step (203), where underwater image (101) denoising techniques are applied to remove noise artifacts and improve the initial quality of the underwater image (101). Following the said, the underwater image (101) is resized in step (204) to ensure uniformity and compatibility with the hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model used in subsequent stages.

[043] The system continues its operation by processing the resized images through the Image Enhancement Module (102). Step (205) includes applying White Balance to correct color distortions caused by the absorption of red wavelengths in underwater environments and Contrast Limited Adaptive Histogram Equalization (CLAHE) to enhance the contrast of the underwater image (101), improving visibility of fine details. In step (206), the Dark Channel Prior Algorithm is employed to further reduce haze, enhancing clarity by leveraging statistical properties of haze-free underwater image (101).
[044] Once the underwater image (101) is enhanced, the system performs Non-Anomalous Information (NAI) Identification in step (207). The said ensures the identification of key features within the images that are relevant for classification and restoration tasks. The enhanced images are then split into two subsets: training samples in step (208) and testing samples in step (209).

[045] The training samples are forwarded to the Object Classification and Restoration Module (103), where the hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model is trained in step (210). The said hybrid model leverages the feature extraction capabilities of the Convolutional Neural Network (CNN) and the sequential data processing capabilities of the Long Short-Term Memory (LSTM) network to identify and classify objects such as fish, corals, underwater structures, or debris. The training process optimizes the model's parameters for accurate object classification and restoration.

[046] Testing samples, on the other hand, are used in step (212) to evaluate the trained hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) model. The said step (212) ensures that the trained model performs reliably on unseen data. Once the testing is complete, the system generates enhanced underwater images with classified objects in step (211). These final outputs are visually improved versions of the underwater images with labeled objects, making them suitable for applications like marine exploration, biodiversity studies, and environmental monitoring.

[047] The system further conducts performance analysis in step (214), where metrics such as accuracy, Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index Measure (SSIM) are evaluated to assess the effectiveness of the system. Finally, in step (215), detailed results, including performance metrics, are compiled into reports that provide a comprehensive assessment of the underwater image classification, restoration, and enhancement system (100).
[048] As per the illustrated and preferred embodiment, the underwater image classification, restoration, and enhancement system (100) operates efficiently by integrating preprocessing techniques, enhancement algorithms, and deep learning methodologies to address the specific challenges of underwater imaging. The system ensures the accurate enhancement, classification, and restoration of underwater images, facilitating real-world applications in marine and environmental research.

TECHNICAL ADVANCEMENTS
[049] The present invention offers the following technical advancements over conventional underwater image restoration system (100):
• The invention utilizes advanced image enhancement techniques, such as white balancing, CLAHE, and Dark Channel Prior, to significantly improve the visual quality of underwater images. By combining these techniques with a Hybrid CNN-LSTM deep learning model, the system restores colour balance, enhances contrast, and reduces visual noise on-the-fly, ensuring clarity and consistency in varying underwater environments.
• Unlike traditional methods that focus solely on spatial details, the proposed hybrid CNN-LSTM model leverages the strengths of Convolutional Neural Networks (CNNs) for spatial feature extraction and Long Short-Term Memory (LSTM) networks to capture sequential and contextual relationships across frames. This combination allows for improved noise reduction, colour restoration, and super-resolution, resulting in consistent image quality and reduced distortions.
• The LSTM network is trained on large datasets of both paired and unpaired underwater images, allowing the model to adapt to different lighting conditions, water turbidity, and depths. This adaptability ensures the system delivers consistent performance across a wide range of underwater scenarios, making it suitable for diverse applications.
• The invention incorporates multi-scale feature extraction through the hybrid CNN-LSTM architecture, allowing for accurate detection and classification of objects ranging from small marine species to large underwater structures. This capability enhances the system’s utility for tasks like marine life monitoring and underwater exploration.
• The invention integrates real-time image enhancement and object classification into a unified workflow, allowing images to be processed and classified simultaneously. This ensures minimal latency and maximizes accuracy, making the system highly suitable for underwater missions requiring quick decision-making.
• The system is adaptable across various underwater industries, including marine biology, archaeological exploration, and environmental monitoring. Its real-time capabilities and advanced image processing ensure reliable monitoring, detection, and classification of marine life, underwater artifacts, and ecological changes.

[050] The foregoing disclosure has been described with reference to the accompanying embodiments which do not limit the scope and ambit of the disclosure. The description provided is purely by way of example and illustration.

[051] The embodiments herein and the various features and advantageous details thereof are explained with reference to the non-limiting embodiments in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.

[052] The foregoing description of the specific embodiments so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the embodiments as described herein.

[053] Any discussion of devices, articles or the like that has been included in this specification is solely for the purpose of providing a context for the disclosure. It is not to be taken as an admission that any or all of these matters form a part of the prior art base or were common general knowledge in the field relevant to the disclosure as it existed anywhere before the priority date of this application.

[054] While considerable emphasis has been placed herein on the components and component parts of the preferred embodiments, it will be appreciated that many embodiments can be made and that many changes can be made in the preferred embodiments without departing from the principles of the disclosure. These and other changes in the preferred embodiment as well as other embodiments of the disclosure will be apparent to those skilled in the art from the disclosure herein, whereby it is to be distinctly understood that the foregoing descriptive matter is to be interpreted merely as illustrative of the disclosure and not as a limitation.
, Claims:WE CLAIM:
1. A method for underwater image classification, restoration, and enhancement, the method comprising the steps of:
- acquiring underwater images (101) from one or more imaging sources, including underwater cameras, remotely operated vehicles (ROVs), autonomous underwater vehicles (AUVs), or sonar imaging systems;
- preprocessing the acquired underwater images (101) using enhancement techniques, including noise reduction, resizing, normalization, histogram equalization, and contrast-limited adaptive histogram equalization (CLAHE);
- training a plurality of deep learning models using annotated underwater images (101) with the application of one or more deep learning techniques, including Convolutional Neural Networks (CNN);
- validating and testing the trained deep learning models with a separate dataset of underwater images (101) to evaluate performance metrics, including accuracy, precision, recall, and F1-score;
- deploying the validated deep learning models into a practical underwater imaging system for real-time application; and
- utilizing the deployed deep learning models to classify objects, enhance image visibility, and provide actionable insights from the underwater images (101).

2. The method as claimed in claim 1, wherein the long short-term memory (LSTM) is trained on paired and unpaired underwater images (101) to learn the transformations required for noise reduction and colour correction, thereby enabling the image enhancement module (102) and Object classification & restoration module (103) to adapt to various underwater conditions, including different lighting and turbidity levels.

3. The method as claimed in claim 1, wherein the training of the deep learning models includes data augmentation techniques to improve model generalizability and robustness.

4. The method as claimed in claim 1, wherein the deployment integrates the deep learning models into cloud-based, edge, or standalone systems for real-time underwater imaging applications.

5. The method as claimed in claim 1, wherein the actionable insights generated include object classification (104), habitat monitoring, infrastructure inspection, and navigation support.

6. A system (100) for underwater image classification, restoration, and enhancement, the system comprising:
- an acquisition unit (100a) configured to acquire underwater images (101) from one or more imaging sources, including underwater cameras, ROVs, AUVs, or sonar imaging systems;
- a pre-processing unit (100b) configured to preprocess the underwater images (101) using enhancement techniques, including noise reduction, resizing, normalization, histogram equalization, and CLAHE;
- a training unit (100c) configured to train a plurality of deep learning models using annotated underwater images (101) with one or more deep learning techniques, including CNNs;
- a validation and testing unit (100d) configured to validate and test the trained deep learning models with performance metrics, including accuracy, precision, recall, and F1-score;
- a deployment unit (100e) configured to deploy the validated deep learning models into real-world underwater imaging applications; and
- an analysis unit (100f) configured to classify objects, enhance image visibility, and generate actionable insights from the underwater images (101).
7. The system (100) as claimed in claim 6, wherein the pre-processing unit (100b) applies colour correction algorithms to mitigate underwater light distortion.
8. The system (100) as claimed in claim 6, wherein the training unit (100c) employs data augmentation techniques to enhance the robustness of the deep learning models.
9. The system (100) as claimed in claim 6, wherein the deployment unit (100e) integrates the deep learning models into a cloud-based, edge, or standalone system for real-time underwater imaging processing.
10. The system (100) as claimed in claim 6, wherein the analysis unit (100f) provides actionable insights, including object classification, habitat monitoring, infrastructure inspection, and navigation support, based on the processed underwater images (101).