Split Summed Area Tables for Blur Processing Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional summed area tables in image processing suffer from memory overflow and rounding errors due to large input values, especially when dealing with large image sizes, which can be resource-intensive to address with higher precision formats.
Innovation Solution
The image processing system segments the input table into multiple sub-tables, calculating output values within each segment independently, reducing the need for higher precision formats and minimizing rounding and overflow errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single summed area table is used for large images, then complete image coverage is achieved, but memory overflow and rounding errors increase
Solution Approach 1:
The patent divides the large image into multiple smaller segments, creating separate summed area tables for each segment. This segmentation approach reduces the memory footprint of each individual table while maintaining complete coverage of the original image. The segments are processed independently and then combined through aggregation to produce the final result, thereby avoiding the memory overflow and rounding errors that would occur with a single large summed area table.
2Measurement precision
If higher precision formats are used to reduce rounding errors, then calculation accuracy improves, but computational resources increase
Solution Approach 1:
By segmenting the image into smaller regions, each with its own summed area table, the patent reduces the accumulation of rounding errors without requiring higher precision data formats. The smaller scope of each segment means that rounding errors accumulate less, maintaining calculation accuracy while using standard precision formats and thus consuming fewer computational resources.
Solution Approach 2:
The patent acknowledges that rounding errors will occur in smaller segments but converts this potential harm into a benefit by designing an aggregation process that combines results from multiple segments. The aggregation step is carefully designed to minimize the impact of individual segment rounding errors, effectively turning the inevitable rounding behavior into an acceptable solution that uses standard precision formats.
Data Source
AI summary
Embodiments are disclosed for performing image processing operations on images using pixel values stored in summed area sub-tables. In particular, in one or more embodiments, the method may include receiving blur parameters for performing a blur operation on an image, the blur parameters including a blur region and a blur size indicating a number of pixels to sample for each output pixel. The method may further include creating a summed area table representation of the image based on pixel values for pixels of the image, the summed area table representation including a plurality of summed area sub-tables. The method may further include performing the blur operation on the blur region of the image using the plurality of summed area sub-tables.