Edge-Preserving Image Upsampling via Vertex Space Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques face challenges in efficiently filtering, upsampling, and colorizing images while preserving edges, particularly in natural scenes containing discrete objects, due to computational intensity and resource requirements.
Innovation Solution
A method involving a computing system that associates pixels with vertices in a vertex space, determining vertex cost matrices based on pixel similarity and spatial/color distances, and generating filtered output images using vertex fit values to maintain edge preservation during image processing tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional image filtering and processing methods are used, then image processing tasks can be completed, but computational resources and time are excessively consumed
Solution Approach 1:
The patent segments the image processing task by identifying and separating edge regions from non-edge regions. By detecting edges first and then applying different processing strategies to edge and non-edge areas, the method reduces unnecessary computations in edge regions while maintaining processing quality, thereby improving productivity and reducing computational resource consumption.
Solution Approach 2:
The patent applies local quality by using edge-aware processing where the filtering strength and computational approach vary based on local image characteristics. In edge regions, minimal filtering is applied to preserve sharpness, while in non-edge regions, more aggressive filtering can be used. This localized adaptation reduces overall computational burden while maintaining image quality.
2Reliability
If aggressive filtering is applied to reduce noise, then noise removal is improved, but edge sharpness and boundaries are degraded
Solution Approach 1:
The patent implements local quality by adapting the filtering strength to local image characteristics. Edge detection identifies regions where sharp transitions occur, and the filtering algorithm adjusts its behavior accordingly: applying minimal filtering at edges to preserve sharpness while applying stronger filtering in homogeneous regions to remove noise. This resolves the contradiction between noise removal quality and edge sharpness.
Solution Approach 2:
The patent segments the image into edge and non-edge regions, applying different filtering strategies to each segment. Edge regions undergo minimal or no filtering to maintain sharpness, while non-edge regions receive aggressive filtering for noise removal. This segmentation approach allows the system to achieve both high noise filtering quality and edge sharpness simultaneously.
3Manufacturing precision
If edge-preserving processing is implemented, then edge sharpness is maintained, but computational complexity increases
Solution Approach 1:
The patent reduces computational complexity through segmentation by first detecting edges and then applying simplified processing to the majority of non-edge regions. The complex edge-preserving algorithms are only applied where necessary (at detected edges), while standard efficient filtering is used elsewhere. This segmentation strategy maintains edge sharpness without requiring complex processing throughout the entire image.
Solution Approach 2:
The patent applies partial action by implementing edge-aware processing that focuses computational effort only where needed. Rather than applying complex edge-preserving algorithms to every pixel, the method identifies edge locations and applies sophisticated processing only in those regions, using simpler and faster processing for the remainder of the image. This partial application of complex algorithms reduces overall computational complexity while maintaining edge preservation quality.
Data Source
Figure 1A~1B
Figure 2A~2B
Figure 2C~3
AI summary
Example embodiments may allow for the efficient, edge-preserving filtering, upsampling, or other processing of image data with respect to a reference image. A cost- minimization problem to generate an output image from the input array is mapped onto regularly- spaced vertices in a multidimensional vertex space. This mapping is based on an association between pixels of the reference image and the vertices, and between elements of the input array and the pixels of the reference image. The problem is them solved to determine vertex disparity values for each of the vertices. Pixels of the output image can be determined based on determined vertex disparity values for respective one or more vertices associated with each of the pixels. This fast, efficient image processing method can be used to enable edge-preserving image upsampling, image colorization, semantic segmentation of image contents, image filtering or de-noising, or other applications.