Weighted Least Squares Image Smoothing Block Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods based on the weighted least squares (WLS) method require significant computational resources, leading to memory and processing inefficiencies.
Innovation Solution
Implementing an image processing device and method that performs one-dimensional smoothing processing in a first direction followed by one-dimensional smoothing processing in a second direction, using a WLS filter, allowing for reduced calculation resources by processing in blocks rather than the entire image at once.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If image smoothing is performed using the weighted least squares method on the entire image at once, then smoothing quality is improved, but memory usage and computational resources increase significantly
Solution Approach 1:
The patent divides the image into multiple blocks and processes each block separately using the weighted least squares method. This segmentation approach maintains the quality of smoothing by applying the algorithm to each block while significantly reducing memory usage, as only one block needs to be held in memory at a time rather than the entire image.
2Manufacturing precision
If image smoothing is performed using the weighted least squares method on the entire image at once, then smoothing quality is improved, but processing time increases
Solution Approach 1:
By segmenting the image into blocks and processing them independently, the patent enables parallel processing of multiple blocks. This maintains smoothing quality while reducing total processing time through concurrent execution on multi-core processors or GPUs.
Solution Approach 2:
The patent performs preliminary calculations of gradient values and weight coefficients for each block before executing the full weighted least squares optimization. This pre-computation reduces the computational complexity during the main smoothing phase, thereby reducing processing time while preserving smoothing quality.
3Manufacturing precision
If one-dimensional smoothing is performed sequentially in both directions on the entire image, then edge preservation is improved, but computational complexity increases
Solution Approach 1:
The patent applies one-dimensional smoothing in horizontal and vertical directions by dividing the image into blocks. This segmentation reduces computational complexity by limiting the scope of each smoothing operation to small local regions while maintaining edge preservation through the weighted least squares method's ability to detect and preserve edges within each block.
Solution Approach 2:
The patent decomposes the two-dimensional smoothing problem into two separate one-dimensional smoothing operations performed sequentially in different directions (horizontal then vertical). This dimensional decomposition reduces computational complexity from O(n^4) to O(n^2) while maintaining edge preservation through the iterative application of 1D smoothing in each direction.
Data Source
AI summary
The purpose of the present invention is to reduce required calculation resources for image smoothing based on weighted least squares. The image processing device 100 includes: a first smoothing unit 110 for applying one-dimensional first smoothing processing based on weighted least squares to an input image in a first direction per a predetermined number of lines; and a second smoothing unit 120 for applying one-dimensional second smoothing processing based on weighted least squares to the first-smoothing-processed input image in a second direction, which is different from the first direction, per a predetermined block.