Lens Correction Lookup Table Decoding With RNP Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional decoders for image processing require complex hardware structures for lens correction, leading to large size and increased costs, and existing data compression methods for lookup tables are inefficient due to high memory requirements and computational complexity.
Innovation Solution
A data processing apparatus with a decoder that uses Recursive Near Polynomial (RNP) methods for decoding and compressing lookup table values, employing a seed memory, gain memory, and calculation block to perform bilinear interpolation and update encoded gain values, reducing hardware size and improving data restoration efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional decoders perform many multiplying operations for lens correction, then decoding accuracy is improved, but hardware size increases and device complexity increases
Solution Approach 1:
The patent replaces multiplying operations with adding and shifting operations in the decoder. This substitution transforms the mechanical/computational system from one requiring complex multiplication circuits to one using simpler addition and bit-shifting circuits, thereby reducing hardware complexity while maintaining decoding accuracy for lens correction
Solution Approach 2:
The patent changes the computational parameters by using Recursive Near Polynomial (RNP) methods that express polynomial multiplications as sequences of additions and shifts. This parameter transformation allows the same decoding accuracy to be achieved with fundamentally different operational parameters that require less complex hardware
2Quantity of substance
If lookup table size is increased to improve data storage capacity, then memory capacity increases, but memory footprint increases
Solution Approach 1:
The patent extracts only the essential information from the full lookup table by using RNP compression to represent the table with a smaller set of coefficients. This extraction allows the system to retain the necessary data storage capacity while removing redundant information that would otherwise occupy memory space
Solution Approach 2:
The patent creates an asymmetric relationship between the compressed representation and the original data. Instead of storing complete symmetric data structures, the system stores asymmetric compressed coefficients that can be algorithmically expanded back to the full data set when needed, reducing memory footprint while preserving data capacity
3Area of stationary object
If polynomial regression coefficients are stored to reduce lookup table memory, then memory footprint is reduced, but computational complexity increases due to large number of multiplications needed for restoration
Solution Approach 1:
The patent replaces the computational mechanism for restoring polynomial data by substituting multiplication operations with addition and shifting operations. This allows the system to maintain reduced memory footprint from coefficient storage while eliminating the computational complexity burden of performing many multiplications during data restoration
Solution Approach 2:
The patent inverts the conventional approach by not directly computing polynomial multiplications during restoration. Instead, it uses RNP methods that work backwards from the coefficients through a sequence of additions and shifts to reconstruct the original data, thereby solving the computational complexity problem while maintaining memory efficiency
4Measurement precision
If more information is transmitted from host to data processing apparatus for precise lens correction, then lens correction precision is improved, but information transmission time increases
Solution Approach 1:
The patent extracts and transmits only the essential correction parameters using RNP compression rather than transmitting complete lookup tables or full correction data sets. This extraction allows precise lens correction to be achieved with a smaller information payload, thereby reducing transmission time while maintaining correction precision
Solution Approach 2:
The patent performs preliminary compression of the correction data into RNP coefficient form before transmission. This preliminary action prepares the data in a compact representation that requires less transmission time while preserving all necessary information for precise lens correction at the receiving end
Data Source
AI summary
A data processing apparatus having a small hardware size and performing lens correction includes a seed memory, a gain memory, and a decoder. The seed memory stores a plurality of encoded gain values forming an A*B seed matrix; the gain memory stores a plurality of decoded gain values forming an M*N grid matrix. The decoder is connected between the seed memory and the gain memory and performs vertical recursive rear polynomial (RNP) decoding before performing horizontal RNP decoding after a second time. Accordingly, the decoder may decode the M*N grid matrix from the A*B seed matrix by performing the row directional RNP decoding and the column directional RNP decoding by turns.