Vector Processor Local Binary Pattern Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Local Binary Patterns (LBP) computation is processor and memory bandwidth intensive due to the need for repeated comparisons of each pixel with its neighboring pixels in image processing, leading to significant memory burdens.
Innovation Solution
The use of a vector processor with multiple memory ports and simultaneous access capabilities to efficiently perform LBP computations by comparing and aggregating binary decisions across multiple pixels in parallel, reducing the number of memory accesses and increasing processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional scalar processors are used to compute LBP by comparing each pixel with its neighboring pixels, then the computation can be performed, but the processor and memory bandwidth are excessively intensive
Solution Approach 1:
The patent segments the LBP computation into multiple independent comparison operations that can be executed in parallel. Each pixel's LBP value is computed by independently comparing it with its neighboring pixels, and these comparisons are organized into vector operations that process multiple pixels simultaneously, reducing the overall memory bandwidth requirement.
Solution Approach 2:
The patent merges multiple comparison operations into single vectorized instructions. By combining the comparison of multiple pixels with their respective neighbors into unified vector operations, the system reduces the total number of memory accesses while maintaining computational accuracy, thereby improving productivity and reducing memory bandwidth usage.
2Measurement precision
If repeated comparisons of each pixel with neighboring pixels are performed using conventional methods, then LBP values are computed, but the number of memory accesses is excessive
Solution Approach 1:
The patent performs preliminary organization of pixel data and comparison operations to minimize memory access time. By pre-arranging data in memory and using vectorized comparison operations that process multiple pixels in parallel, the system reduces the total time spent on memory accesses while maintaining the precision of LBP computations.
3Quantity of substance
If LBP computations are performed on large images, then comprehensive feature extraction is achieved, but the computational burden increases significantly
Solution Approach 1:
The patent replaces traditional scalar mechanical computation with vectorized processing operations. By using vector instructions that can process multiple data elements simultaneously, the system handles large numbers of pixels with reduced computational complexity, enabling efficient LBP computation on large images without proportionally increasing device complexity.
Data Source
AI summary
A method (and system) of determining a local binary pattern in an image includes selecting an orientation. For each pixel in the image, the method further includes determining a binary decision for each such pixel relative to one neighboring pixel of the orientation, selecting a new orientation, and repeating the determination of the binary decision for each pixel in the image relative to one neighboring pixel of the newly selected orientation.