Unlock instant, AI-driven research and patent intelligence for your innovation.

Bayesian image denoising method based on noise-containing image distribution constraint

A Bayesian and image technology, applied in the fields of digital image processing and computer vision, can solve problems such as high cost, difficulty in deep learning, difficulty in obtaining noise-free data, etc.

Pending Publication Date: 2020-11-13
TSINGHUA UNIV
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in most applications, real noise-free data is often difficult or expensive to obtain, which brings difficulties to the application of deep learning in image denoising

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
  • Bayesian image denoising method based on noise-containing image distribution constraint
  • Bayesian image denoising method based on noise-containing image distribution constraint
  • Bayesian image denoising method based on noise-containing image distribution constraint

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0110] Example 1: Denoising Handwritten Digital Image Containing Noise

[0111] The imaging system applied in this method is: handwritten digital acquisition system

[0112] The noise-containing image is: noise-containing handwritten digital image, image resolution 28*28

[0113] The noise-free image is: noise-free handwritten digital image, image resolution 28*28

[0114] (1) According to the distribution of noise-containing handwritten digital images, calculate the distance cost function of the distribution of noise-free handwritten digital images;

[0115] (1-1) Denote the random vector of the handwritten digital image containing noise generated by the handwritten digital acquisition system as Z, and estimate the probability density function f of the random vector Z of the handwritten digital image containing noise in an analytical form Z :

[0116] (1-2) Collect 70,000 noisy handwritten digital images from the handwritten digital acquisition system, and obtain the probabi...

Embodiment 2

[0130] Embodiment 2: Denoising of camera night photography images:

[0131] The imaging system applied in this method is: video camera

[0132] The noise-containing image is: noise-containing photographic image block, image resolution 512*512

[0133] The noise-free image is: noise-free photographic image block, image resolution 512*512

[0134] The resolution of the original noisy photographic image is usually greater than 512*512, and 512*512 noise-containing photographic image blocks can be gradually intercepted from the original noisy photographic image, and each noise-containing photographic image block is denoised to obtain an estimated noise-free photographic image block , the noise-free photographic image blocks are stitched together to obtain an estimated noise-free photographic image.

[0135] (1) Calculate the distance cost function of the noise-free photographic image block distribution according to the noise-containing photographic image block distribution, wher...

Embodiment 3

[0157] Embodiment 3: X-ray CT projection image

[0158] The imaging system applied in this method is: X-ray CT

[0159] The noise-containing image is: noise-containing projection image, image resolution 1024*720

[0160] The noise-free image is: noise-free projection image, image resolution 1024*720

[0161] (1) Calculate the distance cost function of the distribution of the noise-free projection image according to the distribution of the noise-containing projection image, in which the distance cost function of the distribution of the noise-free projection image is estimated using an indirect distribution distance constraint method. The specific steps are as follows:

[0162] (1-1) For a set of noise-containing projection image samples z that need to be denoised by this method 1 ,z 2 ,z 3 ,...,z n Perform Bayesian iterative denoising to obtain estimated noise-free projection image block samples the initial value of .

[0163] (1-2) The noise contained in the noise-cont...

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 belongs to the technical field of digital image processing and computer vision, and particularly relates to a Bayesian image denoising method based on noise-containing image distributionconstraints. According to a Bayesian posterior probability theory, estimation of a noiseless image from a noisy image depends on modeling of prior distribution of the noiseless image. The method comprises the following steps: firstly, under the condition that an additive noise distribution model is known, learning from a noise-containing image sample to obtain the distribution of a noise-free image; according to the method, the constraint of noise-free image prior distribution is converted into the constraint of noise-containing image prior distribution for Bayesian denoising, and the Bayesian denoising implementation method of the image denoising neural network is trained in an unsupervised mode. According to the method, noise-free image implicit distribution characteristics can be accurately learned by fully utilizing noise-containing image samples, so that efficient image denoising is realized.

Description

technical field [0001] The invention belongs to the technical fields of digital image processing and computer vision, and in particular relates to a Bayesian image denoising method based on distribution constraints of images containing noise. Background technique [0002] In the process of image acquisition such as photography, public security monitoring, medical imaging, and microscopic imaging, due to factors such as random characteristics of physical signals, environmental interference, and state changes of the system itself, the collected images inevitably contain noise, which significantly affects Image Quality. The image denoising method aims to reduce the random noise components in the image and restore the effective information of the image, which is an important method and technology in image processing. [0003] Classic image denoising methods include mean filtering, median filtering, Gaussian low-pass filtering and other filtering methods. These filters limit th...

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): G06T5/00G06N3/04G06N3/08G06K9/62G06F17/15G06F17/18G06F17/14
CPCG06N3/08G06F17/15G06F17/18G06F17/14G06N3/045G06F18/22G06T5/70G06T2207/20076G06T2207/20081G06T2207/20084G06N3/084G06N3/047G06N3/048G06T5/60
Inventor 邢宇翔张丽高河伟邓智
Owner TSINGHUA UNIV