Image De-blurring via Salient Edge Kernel Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image de-blurring methods face challenges in accurately estimating the blur kernel and achieving high-quality deconvolution, particularly due to issues with noise removal and edge selection, leading to incomplete de-blurring and artifacts in restored images.
Innovation Solution
The method involves marking edge and smooth regions in an input image to obtain an intermediate image, extracting salient edges to estimate the blur kernel, and performing non-blind deconvolution using hyper-Laplacian and TV-L2 models to restore the image, effectively removing noise and maintaining edge details.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional de-blurring algorithms are used to restore clear image from blurred image, then the restoration process can be completed, but the accuracy of blur kernel estimation is insufficient and artifacts remain in the restored image
Solution Approach 1:
The patent segments the image processing into distinct stages: edge detection, salient edge selection, blur kernel estimation, and non-blind deconvolution. By dividing the restoration process into modular components, each stage can be optimized independently, improving overall accuracy while maintaining reliability.
Solution Approach 2:
The patent performs preliminary edge detection and salient edge selection before blur kernel estimation. This preliminary action prepares the data in advance, ensuring that only the most relevant edge information is used for kernel estimation, thereby improving accuracy without compromising the final restored image quality.
2Measurement precision
If edge information is extracted for blur kernel estimation, then the estimation accuracy can be improved, but noise and ringing artifacts are introduced in the restored image
Solution Approach 1:
The patent applies different processing strategies to different regions of the image. Salient edges are identified and treated with specialized constraints that preserve their structural information for accurate kernel estimation, while smooth regions are processed with noise-suppressing constraints. This local differentiation improves estimation accuracy without introducing artifacts in noise-prone areas.
Solution Approach 2:
The patent changes the constraint parameters used in different processing stages. During edge-based kernel estimation, constraints are relaxed to preserve edge details, while during the final restoration phase, stronger regularization is applied to suppress noise and ringing artifacts. This dynamic parameter adjustment resolves the contradiction between accuracy and artifact suppression.
3Reliability
If non-blind deconvolution is performed to restore the input image, then the de-blurring effect can be achieved, but incomplete de-blurring and artifacts remain due to inaccurate blur kernel estimation
Solution Approach 1:
The patent implements an iterative feedback mechanism where the blur kernel estimation is refined based on the results of non-blind deconvolution. The restored image quality is evaluated, and the kernel estimation is adjusted accordingly. This feedback loop continuously improves kernel accuracy, enabling complete de-blurring while minimizing artifacts in the final restored image.
Data Source
AI summary
A method for image de-blurring includes estimating an intermediate image L by marking and constraining an edge region and a smooth region in an input image; estimating a blur kernel k by extracting salient edges from the intermediate image L, wherein the salient edges have scales greater than those of the blur kernel k; and restoring the input image to a clear image by performing non-blind deconvolution on the input image and the estimated blur kernel k. Imposing constraints on the edge region and the smooth region allows the intermediate image to maintain the edge while effectively removing noise and ringing artifacts in the smooth region. The use of the salient edges in the intermediate image L enables more accurate blur kernel estimation. Performing non-blind deconvolution on the input image and the estimated blur kernel k restores the input image to a clear image achieving desired de-blurring effect.