Image Processing Apparatus Super Resolution Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional super resolution technologies enhance noise components when converting low resolution video signals to high definition, particularly when scaling from SD to 4K or full HD, as they interpolate pixels without considering the original signal characteristics.
Innovation Solution
An image processing apparatus comprising a first scaler, a second scaler, a generator, a weight assigner, and a blender, which scales, super-resolves, and blends image signals based on characteristic information to optimize pixel conversion and reduce noise enhancement, using techniques like super resolution conversion and coefficient assignment to generate high-resolution images without emphasizing noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional super resolution conversion is applied to low resolution video signals, then image resolution is improved, but noise components are enhanced
Solution Approach 1:
The patent applies different processing methods to different regions of the image based on local characteristics. The gradient calculation and noise determination are performed on a per-block basis, allowing the system to identify regions with high noise content and apply appropriate filtering or weighting to those specific local areas while preserving detail in cleaner regions.
Solution Approach 2:
The patent dynamically adjusts processing parameters based on the input signal characteristics. By calculating gradients and determining noise levels, the system changes parameters such as blending ratios and filtering strength adaptively, rather than applying fixed super-resolution parameters to all regions, thereby resolving the contradiction between resolution enhancement and noise suppression.
2Measurement precision
If pixel interpolation is used for super resolution conversion, then video resolution is increased, but noise components contained in the input video signal are enhanced
Solution Approach 1:
The patent performs preliminary analysis of the input video signal by calculating gradients and determining noise characteristics before applying the final super-resolution conversion. This preliminary action allows the system to prepare appropriate processing strategies for different regions, identifying areas where noise enhancement is likely to occur and pre-planning mitigation approaches.
Solution Approach 2:
The patent incorporates feedback mechanisms where the results of gradient calculations and noise determinations are used to adjust the super-resolution processing parameters. The system continuously monitors the processed output and adjusts blending ratios and filtering parameters based on the detected noise levels, creating a closed-loop system that prevents excessive noise enhancement while maintaining resolution improvements.
3Device complexity
If the same super resolution conversion is performed on all input video signals regardless of format, then processing simplicity is maintained, but image quality deteriorates due to noise enhancement
Solution Approach 1:
The patent transforms the static, uniform super-resolution processing into a dynamic, adaptive system. The processing parameters, filtering strength, and blending ratios change dynamically based on the input signal characteristics, gradient calculations, and noise determinations. This dynamic approach maintains operational simplicity from the user perspective while internally adapting to optimize image quality for each specific input format and content type.
Data Source
AI summary
According to one embodiment, an image processing apparatus includes a first scaler, a second scaler, a generator, a coefficient assigner, and a blender. The first scaler converts the number of pixels of an image signal to scale up it according to control information set based on characteristic information. The second scaler converts the number of pixels of the image signal scaled up by the first scaler to further scale up it. The generator estimates an original pixel value from the image signal to increase the pixels, and generates an image signal with a resolution higher than that of the image signal scaled up by the first scaler. The coefficient assigner assigns a coefficient to the image signal generated by the generator according to the characteristic information or the control information. The blender blends the image signal generated by the generator with the image signal scaled up by the second scaler.