Image De-noising via Multi-layer Neural Network Feature Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image de-noising methods, such as ID-CNN and depth full-convolution coding-decoding networks, are inadequate in effectively removing noise from images, particularly in scan images, as they fail to accurately distinguish and remove noise across various types and intensities, leading to incomplete noise reduction.
Innovation Solution
A method and device utilizing a multi-layer neural network to create a single dimensional vector, convert it into a multi-dimensional matrix, generate a feature hierarchy, segment the image based on this hierarchy, and remove noise from these segments to produce a de-noised image, employing machine learning techniques for noise detection and validation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional ID-CNN or depth full-convolution coding-decoding networks are used for noise removal, then the de-noising process can be performed, but the noise removal is incomplete and ineffective particularly for scan images with various types and intensities of noise
Solution Approach 1:
The patent segments the image processing task by creating multiple levels in the feature hierarchy, where each level processes specific types of features. The image is divided into segments at different hierarchical levels, allowing targeted noise removal for different feature types while maintaining overall image integrity.
Solution Approach 2:
The patent transforms the traditional 2D image processing into a multi-dimensional feature hierarchy with multiple levels. By organizing features hierarchically across different dimensions, the system can process and remove noise from various feature types simultaneously, achieving more complete noise removal than conventional 2D approaches.
2Measurement precision
If multi-layer neural network with feature hierarchy is used to achieve complete noise removal, then noise removal accuracy is improved, but the processing complexity and computational requirements increase
Solution Approach 1:
The multi-layer neural network is segmented into multiple levels, with each level responsible for detecting and processing specific types of features. This segmentation allows the complex detection task to be distributed across multiple simpler sub-tasks, improving overall detection accuracy while managing computational complexity through hierarchical organization.
Solution Approach 2:
The patent applies partial action by having different levels of the feature hierarchy focus on specific feature types rather than processing all features uniformly. Each level performs specialized processing for its designated feature type, achieving high precision for specific noise types without the computational overhead of exhaustive full-image processing at every level.
3Device complexity
If conventional convolutional neural networks are used for de-noising, then the processing can be performed with simpler structure, but the ability to distinguish and remove various types of noise is insufficient
Solution Approach 1:
The feature hierarchy structure serves multiple functions: it detects different types of features, processes various noise types, and maintains image structure at different scales. This multi-functional design allows a single system to handle diverse noise types (salt-and-pepper, Gaussian, speckle, etc.) while preserving image quality, achieving versatility without requiring separate specialized networks for each noise type.
Data Source
AI summary
A method and device for removing noise from an image is disclosed. The method includes creating a single dimensional vector for an image through a multi-layer neural network. The method further includes converting the single dimensional vector into a multi-dimensional matrix based on number of layers in the multi-layer neural network. The method includes generating a feature hierarchy based on the multi-dimensional matrix, such that the feature hierarchy comprises a plurality of levels, and each level in the plurality of levels comprises at least one feature associated with the image. The method further includes creating a plurality of segments for the image based on the feature hierarchy, such that each of the plurality of segments includes a set of features associated with the image. The method includes removing each segment comprising noise from the plurality of segments to generate a de-noised image.