Ultrasonic Image Filtering Using Partial Matrix Median Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ultrasonic diagnostic apparatuses face challenges in achieving high-speed image processing while maintaining diagnostic accuracy, as median filtering requires a large number of computations, which can be inefficient, especially when using single instruction multiple data (SIMD) processors.
Innovation Solution
The apparatus employs an image processing section that sorts elements on a column-by-column or row-by-row basis, extracts a partial matrix around a reference position, and computes the median of this partial matrix, reducing the number of computations compared to traditional median filters by using a smaller partial matrix.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a traditional median filter is used to smooth boundaries and remove noise, then image quality is improved, but the number of computations increases significantly
Solution Approach 1:
The patent divides the image data into multiple blocks and processes each block independently using parallel computation. By segmenting the processing task across multiple data elements simultaneously, the system maintains high image quality through median filtering while significantly reducing overall processing time through parallel execution.
Solution Approach 2:
The patent applies median filtering selectively to specific regions or uses a reduced computation approach where full sorting is performed only when necessary. This partial application of the filtering operation maintains diagnostic quality in critical areas while reducing unnecessary computations in other regions.
2Manufacturing precision
If the filtering range is increased to improve noise removal, then image quality is improved, but the number of computations increases
Solution Approach 1:
The patent segments the filtering operation into column-wise and row-wise passes, where each pass processes a specific dimension independently. This segmentation allows the system to handle larger effective filtering ranges by breaking down the complex 2D filtering into simpler 1D operations that can be executed more efficiently.
Solution Approach 2:
The patent transforms the 2D median filtering problem into a sequence of 1D filtering operations by processing columns and then rows separately. This dimensional decomposition reduces computation complexity while maintaining the noise removal effectiveness of the original 2D median filter.
Data Source
AI summary
An ultrasonic diagnostic apparatus uses the pixel value of each of the pixels found in a predetermined range relative to a reference position of image data as input matrix and computes an output value corresponding to the input at its image processing section. Thereafter, the sorting filter of the image processing section sorts the elements of each row of the input matrix on a column by column basis. Then, it extracts a partial matrix having the row including the reference position and surrounding rows from the matrix obtained as a result of the sorting and having a number of rows smaller than number of rows of the input matrix and a number of columns same as the number of columns of the input matrix. Furthermore, it computationally determines the median of the elements of the partial matrix and outputs the median as the output value.