Abstract:Artificial retinas have revolutionized the lives of many blind people by enabling their ability to perceive vision via an implanted chip. Despite significant advancements, there are some limitations that cannot be ignored. Presenting all objects captured in a scene makes their identification difficult. Addressing this limitation is necessary because the artificial retina can utilize a very limited number of pixels to represent vision information. This problem in a multi-object scenario can be mitigated by enhancing images such that only the major objects are considered to be shown in vision. Although simple techniques like edge detection are used, they fall short in representing identifiable objects in complex scenarios, suggesting the idea of integrating primary object edges. To support this idea, the proposed classification model aims at identifying the primary objects based on a suggested set of selective features. The proposed classification model can then be equipped into the artificial retina system for filtering multiple primary objects to enhance vision. The suitability of handling multi-objects enables the system to cope with real-world complex scenarios. The proposed classification model is based on a multi-label deep neural network, specifically designed to leverage from the selective feature set. Initially, the enhanced images proposed in this research are compared with the ones that utilize an edge detection technique for single, dual, and multi-object images. These enhancements are also verified through an intensity profile analysis. Subsequently, the proposed classification model's performance is evaluated to show the significance of utilizing the suggested features. This includes evaluating the model's ability to correctly classify the top five, four, three, two, and one object(s), with respective accuracies of up to 84.8%, 85.2%, 86.8%, 91.8%, and 96.4%. Several comparisons such as training/validation loss and accuracies, precision, recall, specificity, and area under a curve indicate reliable results. Based on the overall evaluation of this study, it is concluded that using the suggested set of selective features not only improves the classification model's performance, but aligns with the specific problem to address the challenge of correctly identifying objects in multi-object scenarios. Therefore, the proposed classification model designed on the basis of selective features is considered to be a very useful tool in supporting the idea of optimizing image enhancement.

What problem does this paper attempt to address?

The paper attempts to address the problem of how to improve object recognition capabilities in multi-object scenes within artificial retina systems through image enhancement techniques. Specifically, due to the limited number of pixels in artificial retinas, it is challenging to clearly display all objects in multi-object scenes. Therefore, the paper proposes a classification model based on feature engineering to identify primary objects and filter out non-primary objects, thereby optimizing image enhancement effects and improving visual experience. ### Main Issues 1. **Difficulty in Object Recognition in Multi-Object Scenes**: - The pixel count of artificial retina chips is very limited, making it difficult to clearly represent multiple objects in complex scenes. - Existing simple techniques like edge detection perform poorly in complex scenes, making it hard to accurately identify primary objects. 2. **Need for Image Enhancement**: - A method is needed to enhance images, making primary objects more prominent while removing unnecessary background information. - By optimizing image enhancement, the visual experience of blind patients can be improved, making it easier for them to identify primary objects in scenes. ### Solution The paper proposes a classification model based on deep neural networks (DNN), which selects and extracts key features such as object size, location, and depth through feature engineering to identify primary objects. The specific steps include: 1. **Feature Extraction**: - Use a Single Shot Detector (SSD) model to extract the top 5 objects in the image along with their bounding boxes and prediction scores. - Utilize a monocular depth estimation model to obtain depth information of the image. - Calculate the area and depth of each object to determine its importance. 2. **Classification Model Design**: - The model is divided into two branches: the image branch (i-branch) and the object information branch (o-branch). - The image branch processes image information through 2D convolutional layers and global average pooling. - The object information branch processes object information through 1D convolutional layers and global average pooling. - Finally, the outputs of the two branches are concatenated and passed through a fully connected layer and a sigmoid activation function to output the probability of each object being a primary object. 3. **Experimental Validation**: - Experiments were conducted on the COCO dataset to validate the model's performance on images with single, double, and multiple objects. - Experimental results show that the model's accuracy in identifying primary objects is 96.4% (single object), 91.8% (two objects), 86.8% (three objects), 85.2% (four objects), and 84.8% (five objects). ### Conclusion Through feature engineering and deep learning techniques, the classification model proposed in the paper can effectively identify primary objects and optimize image enhancement, thereby improving the visual experience of artificial retina systems. This method not only solves the problem of object recognition in multi-object scenes but also provides new directions for the future development of artificial retina technology.

Optimizing Image Enhancement: Feature Engineering for Improved Classification in AI-Assisted Artificial Retinas

Artifical intelligence with optimal deep learning enabled automated retinal fundus image classification model

Computer-aided Diagnosis Based on Enhancement of Degraded Fundus Photographs.

Toward Robust and Efficient Low-Light Image Enhancement: Progressive Attentive Retinex Architecture Search

A Hierarchical Image Processing Strategy for Artificial Retinal Prostheses

A Real-Time Image Optimization Strategy Based on Global Saliency Detection for Artificial Retinal Prostheses.

Image Processing Strategies Based on a Visual Saliency Model for Object Recognition under Simulated Prosthetic Vision.

A hybrid model for the detection of retinal disorders using artificial intelligence techniques

Image edge enhancement for effective image classification

OCTNet: A Modified Multi-Scale Attention Feature Fusion Network with InceptionV3 for Retinal OCT Image Classification

An Optimized Content-Aware Image Retargeting Method: Toward Expanding the Perceived Visual Field of the High-Density Retinal Prosthesis Recipients.

Deep Retinal Image Analysis and Classification Using Deer Hunting Optimization-Based Tandem Pulse Coupled Neural Network

A novel approach for automatic classification of macular degeneration OCT images

Enhancing Diabetic Retinopathy Detection Using Pixel Color Amplification and EfficientNetV2: A Novel Approach for Early Disease Identification

A Novel Artificial-Intelligence-Based Approach for Automatic Assessment of Retinal Disease Images Using Multi-View Deep-Broad Learning Network

Automated Identification Of Cataract Severity Using Retinal Fundus Images

Object recognition based on a foreground extraction method under simulated prosthetic vision

An Artificial Intelligence Driven Approach for Classification of Ophthalmic Images using Convolutional Neural Network: An Experimental Study

Enhancing Diabetic Retinopathy Classification: A Fusion of ResNet50 with Attention Mechanism

Classification of retinal image for automatic cataract detection.

Development of an artificial intelligence system to classify pathology and clinical features on retinal fundus images