Image restoration model training method and system and image restoration method
What is Al technical title?
Al technical title is built by PatSnap Al team. It summarizes the technical point description of the patent document.
A technology for repairing models and training methods, which is applied in the field of image repair, and can solve problems such as unnatural areas, difficult repairs, and poor repair effects
Inactive Publication Date: 2020-09-29
CHENGDU UNION BIG DATA TECH CO LTD
View PDF4 Cites 4 Cited by
Summary
Abstract
Description
Claims
Application Information
AI Technical Summary
This helps you quickly interpret patents by identifying the three key elements:
Problems solved by technology
Method used
Benefits of technology
Problems solved by technology
The present invention is aimed at images in complex situations, and aims to solve the problems that the traditional image restoration technology is difficult to repair when the repair area is large and the image color difference is large, the repair effect is not good, and the repaired area is unnatural.
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
Click on the blue label to locate the original text in one second.
Reading with bidirectional positioning of images and text.
Smart Image
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0060] Example 1: Model training
[0061] Step 1 data preprocessing.
[0062] Step 1.1 Specify the number of masks (NUM_MASK), the maximum number of inflection points of the mask (NUM_VER), the minimum distance between two pixels (MIN_LEN), the maximum distance between two pixels (MAX_LEN), two pixels The minimum width of a line segment between points (MIN_BRUSH), the maximum width of a line segment between two pixel points (MAX_BRUSH), and the maximum corner (MAX_ANG) are 3, 5, 50, 200, 10, 40, and 20, respectively.
[0063] Step 1.1 Random mask hyperparameter setting.
[0064] For each image in the dataset, generate an image with the same resolution and a black background. For each image, randomly generate a number from 1 to 3 to represent the number of generated masks. For each mask generation process, a number from 1 to 5 is randomly generated to represent the number of inflection points of the mask.
[0065] Step 1.2 Randomly initialize a pixel within the image resolu...
Embodiment 2
[0086] Example 2: Model Reasoning
[0087] In the model reasoning stage, use the flow as attached Figure 5 shown.
[0088] Step 1 Image collection. In the image restoration stage, users need to select images with objects or targets that are unclear or incomplete according to their own needs.
[0089]Step 2 Prior information. In the prior information stage, the user needs to mark the blurred or incomplete areas in the image on the original image, and the marked area needs to cover the blurred or incomplete areas as much as possible. Afterwards, the user needs to perform a second type of labeling on the edge of the original object in the labeling area. If the edge of the object is blurred or incomplete, users can mark the edge according to their own understanding.
[0090] Step 3 generates mask image and edge information image. In the stage of generating the mask image, according to the user's first type of annotation, the user's annotation area will be drawn on a white b...
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 discloses an image restoration model training method and system and an image restoration method, and the method comprises the steps: carrying out the random mask processing of a plurality of original images, and generating a training data set for training an image restoration deep neural network model; constructing an image restoration deep neural network model, wherein the image restoration deep neural network model comprises a generation model and a discrimination model, the generation model is used for restoring an incomplete area or a fuzzy area of the image, and the discrimination model is used for judging whether a restoration result of the generation model meets a preset requirement or not; training an image restoration deep neural network model by using the training data set to obtain a trained image restoration deep neural network model. According to the method, the image can be restored more effectively, the requirement of people for a high-quality image is met,and the problems that a traditional image restoration technology is difficult to restore under the conditions that the restoration area is large and the image color difference is large, the restoration effect is poor, and the restored area is unnatural can be solved.
Description
technical field [0001] The present invention relates to the field of image restoration processing, in particular to a method and system for training an image restoration deep neural network model based on prior knowledge and an image restoration method. Background technique [0002] Digital image inpainting technology is a technology that uses known information in the image to fill in the defect area in the image, which belongs to the image restoration problem in the field of computer vision. In the field of physical engineering, experts need to analyze the images generated by experiments, but due to the noise of optical devices and other noises, the targets that should be continuous and uniform in the images are broken and weakened. How to restore the lost information in the image as much as possible is an urgent problem to be solved. [0003] Traditional image inpainting methods are mainly divided into two categories, pixel-based image inpainting and block-based image inp...
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
Application Date:The date an application was filed.
Publication Date:The date a patent or application was officially published.
First Publication Date:The earliest publication date of a patent with the same application number.
Issue Date:Publication date of the patent grant document.
PCT Entry Date:The Entry date of PCT National Phase.
Estimated Expiry Date:The statutory expiry date of a patent right according to the Patent Law, and it is the longest term of protection that the patent right can achieve without the termination of the patent right due to other reasons(Term extension factor has been taken into account ).
Invalid Date:Actual expiry date is based on effective date or publication date of legal transaction data of invalid patent.
Login to view more 
Login to view more