Convolution Array Upscaling for Hardware Area Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic apparatuses face challenges in maintaining a small hardware area while performing upscaling operations, particularly for high-resolution image processing, which often results in image quality deterioration due to compression losses during transmission.
Innovation Solution
An electronic apparatus and method that utilize a convolution operation with second kernel data obtained from expanded first kernel data to upscale input data, eliminating the need for additional hardware and reducing hardware area by using a convolution array and shuffler to perform the upscaling operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If additional upscaling layers are added to perform upscaling operations, then the upscaling capability is improved, but the hardware area increases
Solution Approach 1:
The convolution array is designed to perform multiple functions: it can execute standard convolution operations for feature extraction and also perform upscaling operations by selectively applying kernel data. This multi-functionality eliminates the need for separate upscaling layers, thereby maintaining hardware area small while achieving upscaling capability.
Solution Approach 2:
The patent merges the upscaling function with the existing convolution operation in the neural network. By integrating the upscaling capability into the convolution array through selective use of kernel data, the design combines multiple functions (feature extraction and upscaling) into a single hardware component, avoiding additional hardware area consumption.
2Productivity
If compression is applied during transmission of high-resolution images, then transmission efficiency is improved, but image quality deteriorates
Solution Approach 1:
The patent uses kernel data as a template or copy that can be selectively applied during the convolution operation. By storing kernel data (which includes upscaling information) and applying it through convolution, the system reconstructs high-resolution image details from compressed low-resolution input, effectively copying the missing high-frequency information without requiring transmission of the full high-resolution image.
Data Source
AI summary
An electronic apparatus for executing artificial intelligence algorithm is provided. The electronic apparatus includes a memory which stores input data and a plurality of second kernel data obtained from first kernel data, and a processor which obtains upscaled data in which at least a portion of the input data is upscaled by the first kernel data. The data is upscaled by performing a convolution operation on each of the plurality of second kernel data with the input data. Each of the plurality of second kernel data includes a different first kernel element from among a plurality of first kernel elements in the first kernel data.