Low-Resolution Image Conversion Using Patch Classification and Pre-Stored Kernels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for converting low-resolution images to high-resolution images require interpolation processes and large numbers of operations, making them difficult to implement in hardware with low complexity, and often fail to utilize characteristic image information like edges and textures effectively.
Innovation Solution
An image conversion method that divides low-resolution images into patches, classifies them based on characteristics such as edge direction and texture, and uses pre-stored conversion kernels to generate high-resolution patches without interpolation, allowing for efficient conversion and high-quality image generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If interpolation process and floating point conversion kernel are used to convert low-resolution image to high-resolution image, then image quality is improved, but computational complexity and number of operations increase significantly
Solution Approach 1:
The patent divides the low-resolution image into multiple small patches (e.g., 2x2, 4x4 pixels) and processes each patch independently through classification and conversion. This segmentation reduces the computational burden compared to processing the entire image at once, while maintaining high output resolution. Each patch is converted using optimized kernels tailored to its content category.
Solution Approach 2:
The patent performs preliminary classification of image patches into categories (e.g., texture, edge, smooth region) before conversion. This preliminary action enables selection of optimal conversion kernels for each patch type, avoiding unnecessary computations and reducing overall complexity while preserving image quality.
2Ease of manufacture
If conventional interpolation methods are used for image conversion, then implementation is straightforward, but characteristic image information such as edges and textures is not utilized effectively
Solution Approach 1:
The patent applies different conversion kernels to different regions of the image based on their characteristics. Texture regions use one type of kernel, edge regions use another, and smooth regions use a third. This local quality approach preserves edges and textures effectively while maintaining implementation simplicity through automated classification.
Solution Approach 2:
The patent changes the conversion parameters (kernel selection) based on the content characteristics of each patch. By dynamically adjusting which kernel is applied to each patch based on its category, the system optimizes image quality without complex manual intervention.
3Manufacturing precision
If floating point conversion kernel is applied repeatedly to improve image quality, then high-resolution image with high image quality is generated, but number of operations increases significantly
Solution Approach 1:
The patent segments the conversion process into discrete stages: patch extraction, classification, and kernel application. This segmentation eliminates the need for repeated iterative conversions, achieving high quality in a single pass through optimized patch-level processing.
Solution Approach 2:
The patent uses pre-computed conversion kernels that are optimized for specific image categories. These kernels are copied and applied directly to corresponding patches, avoiding the need for repeated learning or iterative improvement processes while maintaining high output quality.
Data Source
AI summary
An image conversion method includes: dividing a low-resolution image to generate a plurality of low-resolution image patches having N (N being a natural number of 2 or greater) number of pixels; classifying the low-resolution image patches into a plurality of image categories; converting the low-resolution image patches by using a conversion kernel for each image patch corresponding to its image category among prestored conversion matrices, to generate high-resolution image patches having L (L being a natural number) number of pixels, which is less than the N number of pixels; and arranging the high-resolution image patches to generate a high-resolution image. According to the image conversion method, an operation amount can be reduced compared with a conventional technique, and thus the image conversion method can be implemented in hardware having low complexity. A high quality and high-resolution image can be generated regardless of characteristic information of an image signal.