Deep Learning Image Reconstruction via Segmented Network Modules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image interpolation methods, such as nearest neighbor, bilinear, and bicubic interpolation, struggle to produce high-resolution images effectively, often resulting in artifacts like jagged edges and blurring, while also being computationally expensive.
Innovation Solution
The proposed image reconstruction method involves a deep learning-based approach using a target reconstruction model that successively invokes a convolution module, a residual network module, and a densely connected network module to progressively enhance image resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bicubic interpolation method is used to improve image resolution, then edge details are preserved effectively, but computational amount increases and data processing speed decreases
Solution Approach 1:
The image reconstruction process is divided into multiple stages with different network modules (convolution module, residual network module, densely connected network module), where each module processes the image at different resolution levels. This segmentation allows the system to achieve high resolution while controlling computational complexity by processing only critical regions at high detail levels.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image. The residual network module focuses computational resources on edge and detail regions that require high precision, while the densely connected network module handles regions requiring moderate detail. This local quality approach preserves edge details effectively while reducing overall computational burden compared to uniform bicubic interpolation.
2Device complexity
If simple interpolation operation is used for image reconstruction, then computational complexity is reduced, but the reconstructed image appears smooth and blurring with low resolution
Solution Approach 1:
The convolution module performs preliminary processing on the input low-resolution image to extract basic features and generate an initial reconstruction. This preliminary action prepares the image data in a form that requires less intensive processing in subsequent modules, achieving a balance between computational complexity and final image resolution.
Solution Approach 2:
The residual network module acts as an intermediary between the convolution module and densely connected network module. It processes the intermediate results from the convolution module, refines edge details, and passes improved features to the densely connected network module. This intermediary processing stage prevents the image from becoming overly smooth and blurring while controlling overall computational complexity.
3Ease of operation
If nearest neighbor interpolation method is used, then computational simplicity is maintained, but artifacts with jagged edges and blurring images are produced
Solution Approach 1:
The patent replaces the simple mechanical interpolation operations of nearest neighbor method with a deep learning-based neural network system. The convolution module, residual network module, and densely connected network module work together to learn complex image patterns and relationships, producing high-quality reconstructions without the jagged edges and blurring artifacts characteristic of nearest neighbor interpolation, while maintaining reasonable computational efficiency through the modular architecture.
Data Source
AI summary
The present application provides an image reconstruction method, a device, equipment, a system, and a computer-readable storage medium. Said method comprises: obtaining a target reconstruction model (S1); invoking a first convolutional layer in the obtained target reconstruction model to extract shallow layer features from the obtained image to be reconstructed (S2); invoking a residual network module in the target reconstruction model to obtain middle layer features from the shallow layer features (S3); invoking a densely connected network module in the target reconstruction model to obtain deep layer features from the middle layer features (S4); and invoking a second convolutional layer in the target reconstruction model to perform image reconstruction on the deep layer features so as to obtain a reconstructed image of the image to be reconstructed (S5). Said method improves the quality and resolution of a reconstructed image.