Electronic Ink Vector Data Denoising via Iterative Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vector-graphic data processing techniques face challenges in denoising and compressing electronic ink data, leading to shape distortion and residual noise, especially in cursive writing, due to overlapping spectra of coordinate signals and hardware-specific noise, which results in loss of critical fragments and aesthetic distortions.
Innovation Solution
The proposed solution involves an iterative global filtering and piecewise reparameterization process in the time (space) domain, using a 3-tap triangular filter for denoising and size reduction, followed by interpolation filtering for reconstruction, which adapts the parametric representation of curves to minimize distortions and noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If a low-pass digital filter is applied to denoise vector-graphic data, then noise is reduced, but shape distortion occurs and critical fragments are lost
Solution Approach 1:
The patent divides the curve into multiple segments and applies different processing strategies to each segment. Critical points are identified and preserved, while non-critical points are subjected to filtering. This segmentation allows noise reduction in less critical areas while maintaining shape integrity in important regions.
Solution Approach 2:
The patent applies different levels of filtering to different parts of the curve based on their importance. Critical fragments and sharp features are protected from heavy filtering, while smoother, less critical segments undergo more aggressive denoising. This local differentiation enables selective noise reduction without compromising overall shape accuracy.
2Quantity of substance
If sample points are removed for compression, then data size is reduced, but quality and legibility deteriorate
Solution Approach 1:
The patent extracts and removes only the least significant sample points while preserving critical points that define the curve's essential features. By identifying and retaining only the necessary points for maintaining legibility and shape, the method achieves compression without sacrificing quality.
Solution Approach 2:
The patent changes the parameterization of the curve by reparameterizing it based on critical points and segments. This allows the same visual quality to be achieved with fewer sample points, as the curve is represented more efficiently through its key features rather than uniform sampling.
3Object-affected harmful factors
If filtering is applied to reduce noise, then denoising is improved, but residual noise and shape distortion remain due to overlapping spectra
Solution Approach 1:
The patent performs preliminary identification and isolation of critical points before applying filtering. By pre-marking essential features, the subsequent filtering process can target only non-critical regions, reducing residual noise while preserving the original signal's essential characteristics.
Solution Approach 2:
The patent uses feedback from curve analysis to adjust the filtering process. By continuously monitoring the curve's geometric properties and comparing them against critical features, the method adapts the filtering strength and characteristics to minimize both noise and distortion.
Data Source
AI summary
In one embodiment, a method for encoding source data includes defining a source polygon using a parametric representation associated with the source data, and identifying one or more portions of the source polygon. The method further includes passing a filter throughout the entire sequence of the samples representing the source polygon, and performing a piecewise, portion by portion, re-parameterization of the source polygon using the corresponding portions of the filtered polygon to obtain an updated sequence of samples representing the source polygon. The passing of the filter and the performance of the piecewise re-parameterization are then iteratively repeated until a convergence condition is satisfied, to provide encoded data.