Image Processing Apparatus Using Weighted Representative Filters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods require large memory capacity and high operation amounts due to the need to store and apply distinct filters for each pixel, and linear interpolation methods fail to generate intermediate filters with high precision when filter characteristics do not change with distance.
Innovation Solution
An image processing apparatus and method that holds representative filters and calculates weight vectors for each pixel, allowing the representative filters to act on the image and combine results weighted by these vectors, reducing the need for extensive memory and operation resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all filters different between respective pixels are held, then filtering precision is improved, but memory capacity requirement increases
Solution Approach 1:
The patent merges multiple pixel-specific filters into a smaller set of representative filters that can be shared across multiple pixels. Instead of storing separate filters for each pixel, the system stores representative filters and combines them using weight vectors to approximate the effect of pixel-specific filters, thereby reducing memory capacity while maintaining filtering precision.
Solution Approach 2:
The patent creates approximate copies of pixel-specific filters by linearly combining representative filters using weight vectors. Rather than storing exact copies of all pixel-specific filters, the system generates approximate filter responses through weighted combinations of representative filters, reducing memory requirements while preserving essential filtering characteristics.
2Quantity of substance
If linear interpolation is used to generate filters between representative filters, then data amount of filters is reduced, but operation amount increases
Solution Approach 1:
The patent performs preliminary calculation of weight vectors that represent the optimal combination of representative filters for each pixel. These weight vectors are pre-computed based on the relationship between representative filters and target pixel filters, allowing the system to avoid complex real-time interpolation operations during actual filtering, thereby reducing operation amount while maintaining low data storage requirements.
3Quantity of substance
If representative filters are held and linear interpolation is used, then memory capacity is reduced, but filtering precision deteriorates
Solution Approach 1:
The patent changes the parameter representation from storing complete filter kernels for each pixel to storing compact weight vectors that define linear combinations of representative filters. This parameter transformation allows the system to maintain filtering precision by accurately representing pixel-specific filter characteristics through weighted sums of representative filters, while significantly reducing memory capacity requirements.
Data Source
AI summary
A plurality of representative filters are held, and weight vectors containing weight values for the respective representative filters as components are acquired for respective pixels which form an image. The respective representative filters act on the respective pixels which form the image, and the results of the action are weighted with the weight vectors and added.