Image Processing Apparatus Feature Vector Quality Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques face challenges in adjusting image quality and encoding amount while reducing calculation costs, as they often result in loss of image details and limited applicability due to reliance on pre-processing methods like filtering, which are costly in terms of computation and not universally effective.
Innovation Solution
An image processing device that generates feature vectors based on pre-decided filters and updates pixel values to optimize quality and encoding amount through quality and encoding amount evaluations, using feedback vectors to determine updating amounts, thereby reducing calculation costs and maintaining image details.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If pre-processing filtering is applied to remove high frequency components, then encoding amount is reduced, but image details and quality deteriorate
Solution Approach 1:
The patent applies filtering as pre-processing before encoding to remove high frequency components that require large encoding amounts. This preliminary action reduces the encoding burden while preserving essential image information through careful filter design that maintains image quality.
2Measurement precision
If conventional quality evaluation and encoding amount evaluation are performed separately with re-encoding, then adjustment accuracy is improved, but calculation cost increases
Solution Approach 1:
The patent merges quality evaluation and encoding amount evaluation into a unified process that operates on feature vectors extracted from the image. By combining these evaluations and avoiding repeated full encodings, the system achieves accurate adjustment while significantly reducing calculation costs.
Solution Approach 2:
Instead of performing full re-encoding to evaluate quality and encoding amount, the patent uses feature vector extraction as a simplified copy or approximation of the full encoding process. This allows evaluation without the computational burden of complete re-encoding while maintaining sufficient accuracy for adjustment decisions.
3Quantity of substance
If adaptive filtering is applied to remove noise components, then encoding amount is reduced, but image details are lost and applicability is limited
Solution Approach 1:
The patent applies filtering with carefully designed parameters that adapt to local image characteristics. By adjusting filter strength and type based on local features, the system removes noise and high frequency components that increase encoding amount while preserving important image details and maintaining broad applicability across different image types.
Data Source
AI summary
An image processing device that updates a pixel value of a processing target image and generates a new image generates a first feature vector based on the processing target image and a first feature map generated with at least one pre-decided filter; updates the processing target image to generate an updated image; generates a second feature vector based on the updated image and a second feature map generated with at least one pre-decided filter; performs quality evaluation of the updated image based on the first and second feature vectors and generates a quality feedback vector which is a vector based on a result of the quality evaluation; performs an encoding amount evaluation on the updated image and generates an encoding amount feedback vector which is a vector based on a result of the encoding amount evaluation; and determines an updating amount in updating of the updated image based on the quality feedback vector and the encoding amount feedback vector.