Integral Image Generation via Area Segmentation and Differential Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for generating integral images require excessive memory storage and hinder parallel processing due to the large memory requirements and dependency of pixel values, making them inefficient for high-quality image processing.
Innovation Solution
An image processing device and method that splits an input image into multiple areas, sets an intersection point as a starting point for generating integral images, and calculates the sum of pixel values in specific areas using these integral images, allowing for parallel processing and reduced memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an integral image is generated by accumulating all pixel values from a reference coordinate, then the sum of pixel values within an arbitrary image area can be calculated with a predetermined number of operations, but the amount of memory required to store the integral image becomes much larger than that for storing the input image
Solution Approach 1:
The patent divides the integral image storage into multiple differential images, where each differential image stores the difference between adjacent pixel values in the accumulation direction. This segmentation transforms the single large integral image into multiple smaller differential images, reducing peak memory requirements while preserving the ability to calculate arbitrary area sums through integral reconstruction.
2Measurement precision
If the integral image stores the sum of pixel values in areas, then the maximum value of stored values becomes much greater than the maximum pixel value, but this increases memory requirements and reduces parallelization capability
Solution Approach 1:
The patent changes the parameter being stored from absolute cumulative sums to differential values (differences between adjacent cumulative sums). This parameter transformation allows the same area sum information to be represented using smaller value ranges that fit within standard pixel data types, reducing memory capacity requirements while maintaining computational accuracy.
3Reliability
If each value of the integral image is affected by the previous value in a sequential accumulation process, then parallelization beyond a certain level becomes impossible, but this reduces processing efficiency for high-quality images
Solution Approach 1:
The patent segments the image processing into multiple independent differential image generation processes that can be executed in parallel. Each differential image can be computed independently from its corresponding input image region, eliminating the sequential dependency that prevents parallelization in traditional integral image generation, thereby improving processing productivity for high-quality images.
Data Source
AI summary
A device for processing an integral image according to the present disclosure may include a memory configured to store at least one process for processing an integral image; and a processor configured to perform an operation according to the process, wherein the processor includes: a generation module configured to split an input image into a plurality of areas, set one of at least one intersection where the plurality of areas intersect as a starting point for generating an integral image, and generate a plurality of integral images for the plurality of areas based on the starting point; and a calculation module configured to calculate a sum of pixel values in a specific area of the input image using the plurality of integral images.