Image super-resolution processing improvement method based on sparse representation
A super-resolution and sparse representation technology, applied in image data processing, graphic image conversion, neural learning methods, etc., can solve the loss of direction information, the effect is weakened, and low-resolution image blocks cannot accurately match high-resolution image blocks. And other issues
Pending Publication Date: 2020-09-01
CHONGQING UNIV
View PDF7 Cites 2 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, a large number of studies have shown that the joint dictionary training method loses the corresponding direction information between low-resolution image blocks and high-resolution image blocks, resulting in the inability of the input low-resolution image blocks to accurately match the corresponding high-resolution image blocks.
As a result, the reconstructed high-resolution image always has artifacts and blurred edges at the edges
Also, the algorithm's effectiveness diminishes rapidly as the required magnification increases
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0027] The following is further described in detail through specific implementation methods:
[0028] This example uses MATLAB 2018a software to simulate the proposed method. The training data set adopts "Yang91", which contains 91 high-resolution images. Implementation steps:
[0029] Step 1. Input: the training data set "Yang 91", and perform downsampling and edge feature extraction on it. The convolution template of the feature extraction operator is:
[0030]
[0031] The convolution template is used to perform convolution operation on the high-resolution image to obtain its edge feature image, and the down-sampling operation adopts bicubic interpolation method to finally generate a training image set P={X,Y,F}, where X={x 1 ,x 2 ,...,x n} is a high-resolution image block set, Y={y 1 ,y 2 ,...,y n} is a low-resolution image block set, F={f 1 ,f 2 ,..., f n} is the edge feature image block set;
[0032] Step 2: Use the joint dictionary training method to obtai...
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
Login to View More Abstract
The invention relates to the technical field of image super-resolution reconstruction, and discloses an image super-resolution processing improvement method based on sparse representation. In order toimprove the reconstruction effect of a sparse model, on the premise that artifacts and edge blurring are reduced, a high-resolution image is accurately reconstructed; the method comprises the following steps: firstly, extracting an edge feature image from a high-resolution image in a training set by utilizing a Laplace operator; obtaining three dictionaries by adopting a joint dictionary trainingmethod, namely, a low-resolution image block dictionary, a high-resolution image block dictionary and an edge feature image block dictionary; in the reconstruction stage, utilizing a low-resolution image block dictionary, a high-resolution image block dictionary and an edge feature image block dictionary to perform reconstruction to obtain an initial high-resolution image and an edge feature image. Compared with a reconstructed edge feature image, the initial high-resolution image is represented as edge blurring and contains artifacts, so that in the optimization stage, the edge feature imageserves as a global constraint to recover the edge of the high-resolution image and reduce the artifacts.
Description
technical field [0001] The invention relates to the technical field of image super-resolution reconstruction, in particular to an improved image super-resolution processing method based on sparse representation. Background technique [0002] At present, the demand for high-resolution images is increasingly urgent in the fields of video surveillance, medical diagnosis and remote sensing applications. In recent years, researchers have proposed a large number of methods to achieve high-resolution image recovery from various degraded images. These methods can be classified into three categories: interpolation-based methods, reconstruction-based methods, and learning-based methods. With the increase of magnification, the reconstruction effect of the interpolation-based method and the reconstruction-based method will decrease rapidly, and the reconstructed image will become too smooth and blurred. The learning-based method is considered to be the most suitable for super-resoluti...
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
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06T3/40G06N3/04G06N3/08
CPCG06T3/4053G06T3/4046G06N3/08G06N3/045
Inventor 周琳朱冰莲
Owner CHONGQING UNIV