Image Rotation via Transposition and Interleaving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image rotation methods, particularly in portable terminals, face inefficiencies when applying Single Instruction Multiple Data (SIMD) technology for 90-degree or 270-degree rotations due to its limitation in processing consecutive data only.
Innovation Solution
The proposed method involves loading each row of pixels into a load memory vector, performing transposition and interleaving operations in a specific sequence, and storing the vectors in reverse order to efficiently rotate 2n×2n pixel images, utilizing a controller with interleaving and transposition units to achieve effective image rotation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If SIMD technology is used for image rotation, then data processing speed is improved, but it cannot effectively process 90 degree or 270 degree image rotation due to its limitation of processing only consecutive data
Solution Approach 1:
The image rotation process is divided into multiple stages: loading consecutive data into SIMD registers, performing transposition operations to rearrange data positions, and executing interleaving operations to achieve final rotation. This segmentation allows SIMD technology to process consecutive data efficiently while achieving 90 or 270 degree rotation through multiple transformation steps.
Solution Approach 2:
Before performing the actual rotation, the method loads data into SIMD registers in a specific order and performs preliminary transposition operations to prepare the data in a state that enables subsequent interleaving operations to complete the 90 or 270 degree rotation. This preliminary preparation allows SIMD to work with consecutive data while achieving non-consecutive rotation results.
2Adaptability or versatility
If general image rotation method is used, then 90 degree or 270 degree rotation can be achieved, but data processing speed is not improved compared to conventional methods
Solution Approach 1:
The patent replaces traditional mechanical image rotation operations with SIMD-based parallel processing operations. By using transposition and interleaving operations in SIMD registers, the method achieves 90 or 270 degree rotation at accelerated speeds compared to conventional sequential processing methods.
3Manufacturing precision
If transposition and interleaving operations are performed multiple iterations, then image rotation accuracy is improved, but processing time increases
Solution Approach 1:
The method performs transposition and interleaving operations for a specific number of iterations (n iterations where n depends on image size) rather than exhaustive processing. This partial action approach achieves sufficient rotation accuracy for practical applications while avoiding unnecessary additional processing time that would result from excessive iterations.
Data Source
AI summary
Provided is an image rotation method and apparatus for rotating an original image of 2n×2n pixels when n is a natural number greater than 1, including loading each row of pixels of the original image into a corresponding load memory vector; and, after the load step, for at least one iteration, performing a transposition operation for each matched load memory vector after matching the load memory vectors and, for zero or more iterations, an interleaving operation between each matched load memory vector after matching the load memory vectors, while the transposition step and the interleaving step are performed a total of n iterations.