A method and system for beautifying portraits

Abstract

The invention discloses a method and system for beautifying portraits, comprising: performing grid division on the image to be processed according to the resolution of the image to be processed to obtain an original grid model; performing face key point detection on the image to be processed , to obtain the key points of the original face; according to the predetermined beauty function, the original grid model is sequentially transformed to obtain the transformed grid model; there are more than one predetermined beauty functions; according to the transformed The mesh model and the original face key points are used to calculate the transformed face key points; the transformed mesh model is rendered and / or according to the transformed face key points, the beauty model is obtained image. The technical solution provided by the invention can perform smooth real-time rendering on the mobile terminal, and has good aesthetic effect.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method and system for beautifying portraits. Background technique [0002] In recent years, beauty selfie camera applications have been popular among users. Such applications usually include functions such as facial skin smoothing, facial liquefaction, beauty makeup, stickers, and filters. Among them, facial liquefaction (also known as portrait beauty) is an important augmented reality attribute of this type of application, which brings many new and interesting ways to play for users, for example, eye reshaping (big eyes, stretched eye corners, eye distance adjustment, etc.), face reshaping (thin face, small face, chin stretch, forehead stretch, etc.), facial features reshaping (wide face / round face / heart-shaped face / diamond face, exaggerated and interesting face shape, etc.), through different The liquefaction method of the portrait face can achieve different...

AI Technical Summary

Problems solved by technology

However, this type of method usually has the following disadvantages: (1) Some beautiful operations need to perform the same operation repeatedly on the basis of the previous operation step, that is, the execution of the next step depends on the result of the previous step. In this case, it is necessary to The entire image performs multiple GPU pipeline renderings, and the final time-consuming also increases linearly with the number of renderings

(2) Part of the beauty function (such as: thin face) cannot obtain the position information of the key points of the deformed face after rendering by executing the shader script, which will affect the subsequent accurate deformation of the face and the sticker function

(3) The shader script will be executed in the pixel shader stage of the GPU, so each pixel of the image will execute the same algorithm and logic code, but in reality, except for the face area, pixels in other areas do not need to execute the corresponding logic code, which is It is bound to further affect rendering performance

(4) The execution of shader scripts is all done by the GPU, making the CPU utilization extremely low

(5) The shader script language is not good at processing loops, branches, bit operations, and logic operations, and some algorithms and logic programming are not easy to implement (such as the feathering mask function of the liquefaction area)

It can be seen that the above disadvantages will affect the execution speed and effect of the beauty model

And when the user opens many beauty functions, the above method also cannot perform smooth real-time rendering on the mobile terminal

Images