Neural Network Image Up-Conversion via Block Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning techniques for image up-conversion, such as convolutional neural networks, face challenges in achieving high-quality image conversion in real-time environments without significant computational overhead, often compromising between processing speed and image quality.
Innovation Solution
A computer system that employs a trained neural network to convert images from a lower resolution to a higher resolution by dividing the image into blocks, adding context data, and applying separable block transforms through multiple layers, allowing for real-time processing and high-quality image generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional convolutional neural networks are used for image up-conversion, then image quality can be improved, but processing speed deteriorates due to significant computational overhead
Solution Approach 1:
The image is divided into multiple blocks, and each block is processed independently through the neural network. This segmentation allows parallel processing of multiple blocks simultaneously, significantly improving processing speed while maintaining image quality. The patent applies this by dividing the input image into N×N blocks and processing each block through separate neural network operations.
Solution Approach 2:
Instead of processing the entire high-resolution image through the neural network, the patent processes only selected blocks (either all blocks or a subset based on importance). This partial action approach reduces computational overhead while still achieving high-quality up-conversion in critical regions, balancing speed and quality requirements.
2Speed
If high-resolution images are processed in real-time, then processing speed is maintained, but computational resources and memory bandwidth increase significantly
Solution Approach 1:
By segmenting the image into blocks, the patent enables selective processing where only necessary blocks are handled in real-time. Less critical blocks can be processed with lower priority or skipped, reducing immediate computational resource usage while maintaining real-time performance for important regions.
Solution Approach 2:
The patent applies different processing quality levels to different blocks based on their importance. High-priority blocks receive full neural network processing for maximum quality, while low-priority blocks use simplified processing or are skipped entirely. This local quality approach optimizes resource allocation to match actual display needs.
3Manufacturing precision
If image blocks are processed through multiple neural network layers, then image quality is improved, but processing time increases
Solution Approach 1:
The patent allows for selective application of multiple neural network layers based on block priority. High-priority blocks can undergo the full multi-layer processing for maximum quality, while lower-priority blocks use fewer layers or single-pass processing. This partial application of excessive action maintains quality where needed while reducing overall processing time.
Data Source
AI summary
A computer system is provided for converting images through use of a trained neural network. A source image is divided into blocks and context data is added to each pixel block. The context blocks are split into channels and each channel from the same context block is added to the same activation matrix. The action matrix is then executed against a trained neural network to produce a changed activation matrix. The changed activation matrix is then used to generate a converted image.