CNN Image Processing via Block Segmentation and Interpolation Lookup
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital image processing methods for convolutional neural networks (CNNs) face inefficiencies in operation workload due to the lack of effective segmentation and interpolation techniques, leading to increased processing time for large input samples.
Innovation Solution
The proposed method segments input image data into feature blocks using an interpolation lookup table, assigns corresponding area interpolations, and calculates interpolation parameters based on multiple training results, allowing for block convolution and efficient combination of output features to reduce processor workload.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the input image data is processed as a whole without segmentation, then the processing is simpler to implement, but the operation workload of the processor increases significantly
Solution Approach 1:
The patent divides the input image data into multiple feature blocks and processes them independently through block convolution. This segmentation reduces the overall operation workload by breaking down a large-scale computation into smaller, manageable units that can be processed in parallel or sequentially with reduced memory bandwidth requirements.
2Productivity
If block convolution is applied to reduce operation workload, then processing efficiency improves, but the device complexity and implementation difficulty increase
Solution Approach 1:
The patent pre-calculates interpolation parameters and stores them in lookup tables before the actual image processing. This preliminary action eliminates the need for complex real-time interpolation calculations during block convolution, significantly reducing the computational complexity while maintaining the benefits of segmented processing.
Solution Approach 2:
The patent introduces interpolation parameters as an intermediary element that facilitates the combination of adjacent feature blocks. These parameters, stored in lookup tables, act as mediators that enable seamless blending of block boundaries without requiring complex coordination logic during the convolution process.
3Measurement precision
If multiple training results are used to calculate interpolation parameters, then the processing precision improves, but the calculation time and operation workload increase
Solution Approach 1:
The patent pre-calculates interpolation parameters from multiple training results and stores them in lookup tables during the training phase. This preliminary computation allows the system to benefit from multiple training results for higher precision without incurring the computational cost during real-time image processing, as the parameters are readily available during inference.
Data Source
AI summary
A system is adapted to perform an image processing method. The processing method includes: obtaining input image data, a first training result, a second training result, and an interpolation lookup table; segmenting the input image data into a plurality of feature blocks according to a total quantity of area interpolations; establishing a position mapping relationship to record the feature blocks corresponding to positions of all of the area interpolations; assigning corresponding area interpolations to the feature blocks according to the position mapping relationship; obtaining an interpolation parameter for each of the feature blocks according to the first training result, the second training result, and the area interpolation; performing block convolution on each of the interpolation parameters and the corresponding feature block to obtain an output feature result; and obtaining an output image by combining the output feature results according to the position mapping relationship.