Integral Image Sub-block Segmentation Memory Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing integral image technique requires a large amount of memory to store accumulated values of entire images, leading to high memory usage, especially as the size of the image increases, and is typically implemented in software due to the large memory requirements.
Innovation Solution
The proposed solution involves dividing input image array data into sub-blocks, generating an integral image for each sub-block, and storing these in separate memory structures, allowing for efficient calculation of the sum of data within a Region of Interest (ROI) by retrieving and combining integral values from multiple memories and buffer units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the integral image technique stores accumulated values of entire images in an individual cell, then the calculation efficiency for ROI sum and image processing (SAD, SSD) is improved, but the memory size required to store the integral image increases significantly
Solution Approach 1:
The patent divides the image data into multiple sub-blocks and generates separate integral images for each sub-block. This segmentation allows the system to store integral values for smaller regions rather than the entire image, significantly reducing the memory size required while still enabling efficient ROI sum calculations by combining results from relevant sub-blocks.
2Adaptability or versatility
If the size of the image (total number of pixels) increases, then the coverage and applicability of the integral image technique is improved, but a larger memory size is required to store the integral image
Solution Approach 1:
By segmenting large images into smaller sub-blocks, the patent enables the integral image technique to handle images of any size without proportionally increasing memory requirements. Each sub-block generates a manageable integral image, and the system adapts to different image sizes by processing appropriate numbers and sizes of sub-blocks.
Solution Approach 2:
The patent introduces a hierarchical dimension by organizing integral images at multiple levels - sub-block level and full image level. This dimensional approach allows efficient handling of large images by working with smaller sub-block integral images when full image processing is not necessary, thus reducing memory requirements while maintaining adaptability to various image sizes.
3Adaptability or versatility
If the integral image technique is implemented in software only due to large memory requirements, then the flexibility and adaptability is improved, but the processing speed is reduced
Solution Approach 1:
The patent enables hardware implementation by segmenting the integral image into sub-blocks, which reduces the memory size to feasible levels. This segmentation makes it practical to implement the technique in hardware (FPGA, ASIC) rather than software only, thereby achieving both hardware-level processing speed and the flexibility to handle different image sizes and configurations.
Data Source
AI summary
An image processing apparatus and method are provided. A control unit may divide input image array data into a plurality of sub-blocks, and a first arithmetic logic unit (ALU) may generate an integral image of at least one of the plurality of sub-blocks. The control unit may determine each of the plurality of sub-blocks to be included in any one of a first sub-block group and a second sub-block group, store the integral image of each sub-block included in the first sub-block group on the first memory, and store the integral image of each sub-block included in the second sub-block group on the second memory.