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

Statistical Compressed Sensing Image Reconstruction Method Based on Hierarchical Gaussian Mixture Model

A Gaussian mixture model and compressed sensing technology, applied in the field of image processing, can solve the problems of non-Gaussian and low-quality reconstructed images, and achieve less measurement, simple iterative form, and good reconstruction effect

Active Publication Date: 2018-08-28
CHINA JILIANG UNIV
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The advantage of this method is: a Gaussian mixture model that describes the statistical properties of the sub-image blocks is formed globally, and the block sparse representation of the sub-image blocks is provided; the disadvantage is that a single Gaussian model is actually used to describe the sub-image blocks locally, but naturally Images usually exhibit significant non-Gaussian properties, and a single Gaussian model cannot flexibly and accurately represent different features existing in sub-image blocks, such as edges, textures, etc., resulting in low quality reconstructed images

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
  • Statistical Compressed Sensing Image Reconstruction Method Based on Hierarchical Gaussian Mixture Model
  • Statistical Compressed Sensing Image Reconstruction Method Based on Hierarchical Gaussian Mixture Model
  • Statistical Compressed Sensing Image Reconstruction Method Based on Hierarchical Gaussian Mixture Model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0022] The present invention will be further described below in conjunction with accompanying drawing.

[0023] refer to figure 1 , the specific implementation process of the present invention is as follows:

[0024] Step 1, divide an image into non-overlapping sub-image blocks, the sub-image block size of this example is ;

[0025] Step 2, for each sub-image block with The sampling rate is compressed and sampled, and the measured :

[0026] ,in is the first sub-image block pixel value, Yes Gaussian random matrix, , is the number of compressed measurements;

[0027] Step 3, generate a value between 0 and 180 degrees Black and white edge images in one direction, calculate all the edge images The covariance matrix of the sub-image patches, generating Gaussian distribution with zero mean direction , using the DCT transform to generate the first direction Gaussian distribution ;

[0028] Step 4, in the first layer of the mixed model, by measuri...

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 statistical compressed sensing image reconstruction method based on a hierarchical Gaussian mixture model, specifically relates to a statistical model learning and signal reconstruction algorithm based on image block data, and mainly solves the existing statistical compressed sensing image reconstruction method. The single Gaussian prior model used in the method cannot flexibly and accurately characterize the non-Gaussian statistical properties of sub-image blocks, resulting in the shortcoming of low reconstructed image quality. This invention learns a global Gaussian mixture model on the first layer of the model to perform hard clustering of sub-image blocks, fully utilizes the similarity structure between image blocks globally, and classifies and learns a local Gaussian mixture model of sub-image blocks on the second layer of the model. , locally distinguish and model the differences between sub-image blocks. Compared with existing traditional compressed sensing and statistical compressed sensing reconstruction technologies, it can obtain higher reconstruction accuracy with fewer measurements, and is suitable for the reconstruction of natural images.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to an image statistical compression sensing reconstruction method, which can be used to reconstruct natural images. Background technique [0002] Compressed Sensing (CS) is a new signal sampling theory formally proposed by Candès and Donoho et al. in 2006, such as: Donoho D L. Compressed sensing. IEEE Transactionson Information Theory, 2006, 52(4): 1289-1306; Candès E. Near optimal signal recovery from random projections: Universal encoding strategies? IEEE Transactions on Information Theory, 2006, 52(12): 5406-525. Different from the traditional sampling theory, CS synchronizes the sampling and compression process, directly perceives the signal in a compressed form, and the number of measurements obtained is much lower than the dimensionality of the sensed signal. In CS theory, under the assumption of a sparse signal model, an exact or approximate reconstructi...

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 Patents(China)
IPC IPC(8): G06T5/00G06T5/50G06T3/40
CPCG06T3/4038G06T5/50G06T2200/32G06T2207/20021G06T2207/20016G06T2207/20081G06T2207/20221G06T5/73
Inventor 武娇曹飞龙
Owner CHINA JILIANG 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