Forward Backward Image Resizing for Noise Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image resolution enhancement methods struggle to effectively filter out noise and defects from low-resolution images, especially when applied to distorted images, and require high computational effort, making them unsuitable for real-time streaming media playback.
Innovation Solution
A forward and backward image resizing method that involves increasing the resolution of a low-resolution image by an integral multiple, followed by a 2-fold increase and then a 2-fold reduction, to produce a high-resolution image while removing noise and maintaining edge structure, requiring low computational effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If common image resolution enhancement methods (interpolation, inverse modeling, training) are applied to low-resolution images, then image resolution is improved, but computational effort and operation time increase significantly
Solution Approach 1:
The patent segments the image processing task into distinct stages: noise filtering using non-local means, forward resizing to intermediate resolution, and backward resizing to target resolution. This segmentation allows each stage to be optimized independently, reducing overall computational complexity while maintaining image quality.
Solution Approach 2:
The patent applies noise filtering using non-local means as a preliminary action before resizing operations. By removing noise and defects early in the process, subsequent resizing operations work with cleaner data, reducing the need for complex iterative optimizations and lowering overall computational effort.
2Manufacturing precision
If noise and defect filtering is performed prior to resolution enhancement, then image quality is improved, but details in the low-resolution image are lost and over-smoothing occurs
Solution Approach 1:
The patent applies non-local means filtering with carefully controlled parameters that preserve local image details while removing noise. The filtering process uses a balance between smoothing strength and detail preservation, applying different levels of processing to different regions based on their characteristics.
Solution Approach 2:
The patent applies a moderate level of noise filtering rather than aggressive filtering, accepting some residual noise to preserve image details. This partial action approach prevents over-smoothing while still improving overall image quality by removing the most prominent defects.
3Manufacturing precision
If forward and backward resizing process is applied, then noise is removed and edge structure is preserved, but processing time increases
Solution Approach 1:
The patent changes the resolution parameter dynamically through forward and backward resizing operations. By temporarily increasing resolution and then reducing it, the process exploits the non-linear relationship between resolution and noise visibility, achieving noise removal and edge enhancement more efficiently than direct resizing.
Data Source
AI summary
A forward and backward image resizing method is used for resizing a low-resolution image into a high-resolution image. In the method, the low-resolution image is obtained first, and then a forward and backward image resizing process is performed, so as to resize the low-resolution image into the high-resolution image with an integral multiple resolution. The forward and backward image resizing process includes: resizing the low-resolution image by the integral multiple, so as to generate a first-resizing image with the integral multiple resolution; further increasing the resolution of the first-resizing image by 2-fold, so as to generate a second-resizing image; and reducing the resolution of the second-resizing image by 2-fold, thereby obtaining the high-resolution image.