Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A general image noise estimation method based on self-similarity measure

A technology of image noise and noise estimation, applied in image enhancement, image analysis, image data processing, etc., can solve the problems of complex algorithm theory, reduced accuracy of image noise estimation, unfavorable application in practice, etc., and achieve the effect of simple theory

Active Publication Date: 2019-03-29
TAISHAN UNIV
View PDF3 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003]The above methods are limited to estimating a single type of image noise, but the noise model in real images often does not only contain a single type of noise, but consists of multiple Mixture of types of noise
The above noise estimation is basically for additive Gaussian noise, but for multiplicative noise, coherent speckle noise, etc., it will often fail; many real images do not have many flat areas, and the noise estimation accuracy for images with fewer flat areas will be reduced. will greatly reduce
These algorithm theories are very complicated, so it is not conducive to the real application in practice
However, in actual image denoising, accurate noise estimation must be performed on the image to effectively achieve image denoising. At present, there is no general and high-precision image noise estimation method.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A general image noise estimation method based on self-similarity measure
  • A general image noise estimation method based on self-similarity measure
  • A general image noise estimation method based on self-similarity measure

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0031] refer to figure 1 As shown, a general image noise estimation method based on self-similarity metric, including the following steps:

[0032] S1. Generate an image with a constant pixel value, and add Gaussian noise with known noise intensity standard deviation to the image to form a standard noisy image, and perform block matching and row matching on the noisy image and calculate the average minimum distance metric D mean , to establish a robust correspondence between the average minimum distance metric and the noise standard deviation through statistical methods;

[0033] S2. Using the corresponding relationship established in S1 to perform noise estimation on the actual noisy image, obtain the noise standard deviation σ1 of the actual noisy image, and simultaneously generate the minimum distance metric mapping image of the actual noisy image;

[0034] S3. Use σ1 in S2 and the generated minimum distance metric mapping image to determine the smooth area in the image, ...

Embodiment 2

[0038] refer to figure 2 as shown, figure 2 Be the method flowchart of the general image noise estimation method based on self-similarity measure of the present invention;'

[0039] The first step: image block matching and row matching calculate the average minimum distance metric;

[0040] refer to image 3As shown, first generate a 256×256 grayscale image with a pixel value of 0.5, the pixel value can also be set as other parameters according to the needs of the image, add Gaussian noise with a noise intensity of standard deviation of 1.0 to the image, and execute the generated For block matching of noisy images, an image block of size N_step is used to extract an image block of size N_step as a reference block, and then block matching is performed in a neighborhood of size NS×NS centered on the reference block to obtain a number of N2 Similar to the image blocks, all the matching image blocks are column-scanned and spliced ​​into a matrix M with a size of (N1×N1)×N2, a...

Embodiment 3

[0052] refer to image 3 As shown, the constant value image and the noisy image in the noise estimation experiment of this embodiment; Figure 4 is an actual noisy image and the distance metric mapping image obtained from it;

[0053] In the research of image denoising in this embodiment, Figure 5 The 10 standard images shown are used MATLAB software to carry out noise addition and noise estimation experiments to the image, adding noise of different intensities to the image each time, and then using the method in Example 2 to noise the image with noise. estimate. Figure 6 is the image noise estimation result of this embodiment. It can be seen from the data in the table that the noise estimation result of this embodiment is very accurate, and can be completely applied to image denoising practice to realize blind image denoising. The estimated results are basically a bit high, because there are more or less noises in the original image.

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a general image noise estimation method based on self-similarity measure, which includes such steps as S1, generating an image with constant pixel value, adding Gaussian noisewith known standard deviation of noise intensity to the image, performing block matching and row matching on the noisy image, and calculating average minimum distance metric Dmean; 2, obtainING that correspondence relationship between the noise standard deviation and the stability of the distance metric according to the mean minimum distance metric Dmean; S3, estimating the noise intensity sigma 1according to the obtained correspondence relation for the actual noisy image and simultaneously generating a distance metric mapping image; S4, judging the smooth region in the image by using the result and the obtained distance metric map image, estimating the noise only in the smooth region of the image, and finally obtaining the accurate noise intensity sigma through two-step iteration. The invention does not need any image transformation, only carries out image self-similarity measurement in the spatial domain, the theory is very simple, and the whole process only uses the Euclidean distance to calculate the self-similarity of the image.

Description

technical field [0001] The present invention relates to the technical field of image noise estimation, and more specifically, to a general image noise estimation method based on self-similarity measure. Background technique [0002] In real life and scientific research, different types of noises with different intensities will inevitably be introduced during the acquisition process of various images. Add noise of known type and known intensity, and then use your own technology to perform image denoising experiments. However, the noise intensity in real images is unknown. In order to achieve effective image denoising, it is necessary to estimate the noise accurately in advance. The existing image noise estimation methods mainly include methods based on wavelet transform and discrete cosine transform, and image noise estimation is realized by counting the transformation coefficients after image transformation. There are two good methods for recent image noise estimation: one...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/223G06T5/00G06T3/40
CPCG06T3/4038G06T7/223G06T5/70
Inventor 侯迎坤侯昊杨洪祥梁凤鸣
Owner TAISHAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com