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

Method for constructing single face skin difference picture pair

A picture-to-picture technology, applied in neural learning methods, image enhancement, neural architecture, etc., can solve problems such as the inability to ensure that the face and facial features are exactly the same

Pending Publication Date: 2022-01-11
杭州缦图摄影有限公司
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The present invention aims at the problems existing in the prior art that the paired pictures constructed by the traditional alignment method have problems in the connection between the non-face area and the area outside the frame, and cannot guarantee that the facial features of the face are exactly the same, and can only be aligned for the part of the face, etc. Defects, providing a new method of constructing a single face skin difference image pair

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
  • Method for constructing single face skin difference picture pair
  • Method for constructing single face skin difference picture pair
  • Method for constructing single face skin difference picture pair

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0047] like Figure 1 to Figure 4 As shown, a method for constructing a single face skin difference image pair includes the following steps:

[0048] S1, take the original picture, the size of the original picture is 512x512 pixels, and the size of the original picture is a multiple of 16, the original picture is copied to obtain the A picture, and then the A picture is smoothed, liquefied, and toned After the operation, the B picture with deformation difference, texture difference, and color difference with the A picture is obtained;

[0049] S2. Construct a deformable grid prediction module a with a convolutional neural network model a, take the original picture in step S1 and input it into the convolutional neural network model a of the deformable grid prediction module a, and take B in step S1 The picture is input into the deformable grid prediction module a, and finally a C picture is obtained;

[0050] S3. Input the C picture obtained in step S2 into the face detection...

Embodiment 2

[0066] In step S1, the size of the original picture is 1024x1024 pixels.

[0067] Other implementations are the same as in Example 1.

Embodiment 3

[0069] In step S1, the size of the original picture is 128x128 pixels.

[0070] Other implementations are the same as in Example 1.

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 method for constructing a single face skin difference picture pair. The method comprises the following steps: copying an original picture to obtain an image A, and performing operating to obtain an image B; inputting the original image into a neural network a, inputting the image B into a grid prediction module a, inputting an obtained image C into a face detection model to construct a face frame to cut the image B and the image C, and a face image A and a face image B are obtained; inputting the face image A into a neural network b, inputting the face image B into a grid prediction module b to obtain an image D, and inputting the image D into a neural network c and a neural network d to obtain a first mask and a second mask; enabling the second mask to correct the first mask to obtain a corrected mask; and obtaining a skin area from the corrected mask by the image D, performing contour feathering, and pasting the skin area back to the image B to obtain an image E, forming an image pair by the image E and the image B, and completing the construction of the single face skin difference image pair. Through the above steps, an image pair of an image E and an image B only having a single face skin difference can be obtained.

Description

technical field [0001] The invention relates to a method for processing pictures, in particular to a method for constructing a pair of single human face and skin difference pictures. Background technique [0002] In the field of portrait beautification, it is often necessary to perform beautification operations on captured pictures. When beautifying, it generally needs to go through liquefaction, skin smoothing and toning operations, so that the original picture can look more beautiful. [0003] Among them, in the operation of skin smoothing and toning, at present, retouchers are usually used to repeatedly skin and toning, which is time-consuming and laborious, and different retouchers have inconsistent retouching results for the same picture. Therefore, with the development of technology With the advancement of the image-to-image deep learning algorithm, the operation of skin smoothing and color correction is not only convenient and quick, but also can obtain the retouchin...

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): G06T3/00G06T5/00G06V40/16G06V40/10G06V10/26G06N3/04G06N3/08
CPCG06N3/084G06N3/045G06T3/04G06T5/94
Inventor 卢海峰
Owner 杭州缦图摄影有限公司