Super Resolution Patch Selection via Feature Vector Classification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing super resolution technologies face challenges in selecting appropriate patch pairs for image restoration, leading to inadequate restoration of blurred images, especially when dealing with diverse types of learning images.

Innovation Solution

An information processing device that includes a proper identifier output unit, a feature vector calculation unit, and a search similarity calculation unit to classify and select proper identifiers for registered patches in a dictionary, enabling accurate similarity calculations for image restoration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If learning images of diverse types are used for super resolution processing, then the adaptability of the restoration system is improved, but the accuracy of patch pair selection deteriorates due to inability to distinguish appropriate patches from inappropriate ones

Engineering Contradiction:

Improveadaptability of restoration systemVSAvoidaccuracy of patch pair selection

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent segments the learning images into different categories based on their content characteristics (e.g., natural images, synthetic images, images with specific objects). By dividing the diverse learning images into segments, the system can selectively apply appropriate patches from relevant segments, improving both adaptability across different image types and accuracy within each category.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary classification mechanism that acts as a mediator between diverse learning images and the super resolution processing. This intermediary classifies and organizes patches from different learning images, enabling accurate selection by determining the appropriateness of each patch for the target image, thus resolving the contradiction between handling diverse types and maintaining selection accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Quantity of substance

If all patch pairs from diverse learning images are processed equally, then the completeness of the dictionary is improved, but the quality of restoration deteriorates due to inclusion of inappropriate patch pairs

Engineering Contradiction:

Improvecompleteness of dictionaryVSAvoidquality of restoration

Core Design Contradiction:

Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent applies local quality by treating different regions and types of patches differently within the dictionary. Instead of uniform processing, the system evaluates each patch pair's appropriateness based on local characteristics (image content, texture, structure) and assigns different weights or selection priorities. This ensures that high-quality appropriate patches are prioritized while maintaining comprehensive coverage through inclusion of diverse but properly weighted patches.

Inventive Principle:

Principle #3Local quality

3Device complexity

If similarity calculation is performed without classification, then the simplicity of the processing algorithm is maintained, but the reliability of patch selection deteriorates due to selection of inappropriate patches

Engineering Contradiction:

Improvesimplicity of processing algorithmVSAvoidreliability of patch selection

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent applies preliminary action by performing classification of learning images and organization of patch pairs before the similarity calculation and super resolution processing. This preliminary classification step organizes the data structure and identifies appropriate patches in advance, making the subsequent similarity calculation more reliable without significantly increasing overall system complexity. The preprocessing ensures that only relevant patches are considered, improving reliability while maintaining algorithmic simplicity.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS9779488B2Information processing device, image processing method and medium

Publication Date: 2017.10.03 NEC CORP

AI summary

An information processing device according to the present invention includes: a proper identifier output unit which outputs proper identifiers for identifying learning images; a feature vector calculation unit which calculates feature vectors of at least a part of patches included in registered patches that are registered in a dictionary for compositing a restored image; and a search similarity calculation unit which calculates a similarity calculation method that classifies the proper identifiers to be given to the registered patches based on the feature vectors.