Example-Based Virtual Fitting Method

Clothing is the basic consumer product that could creates the fashion life.As we all know,clothing also could reflect cultural creativity,technological progress and the times change.Due to the structural characteristics and the shelter function of the clothing,complicated procedures and hidden space are required when people need to change their cloths in daily life,thus causing many inconvenience at shopping.With the development of computer hardware and the increasingly mature of virtual reality technology,people have been able to build virtual characters and virtual costumes,and virtual fitting techniques have been also created.On the one hand,virtual fitting techniques meet the needs of the quick browse of the clothing,on the other hand,it provides technical support for the rapid replacement of virtual characters.Therefore,virtual fitting techniques has been widely used in many fields,such as offline fitting,remote fitting,game entertainment,film and animation.The 3d virtual fitting present the virtual human and clothing in 360 degrees.The key to this technique is the way to dress the virtual human with any type of clothing and simulate the clothing.Our method of fitting could be divided into three steps:firstly,based on human modeling software,clothing design software and clothing animation technology,generating the sample data of human and clothing;Secondly,predict the clothing deformation based on the data-driven method,and deform the standard size sample clothes to fit the target virtual person.Finally,evaluate the fitting effect by quantitative and qualitative analysis.We have achieved the following results:1)A deformation characteristic extraction method based on garment style migration is proposed.The deformation characteristics of this method can make the clothing change with the change of human form and maintain the original design fold feature when deforming.Thus,the users with different body types could have the fitting experience in line with their expectations.2)A data-driven method of fitting simulation is proposed,and based on this method,we simulate the static fitting and dynamic fitting.Firstly,we build the virtual human with different body type and pose base on the characteristics of real human.And then dressing these virtual human through two dimensional garment piece suture technique and clothing animation technique.Secondly,we define a kind of affine transformation between triangular meshes to establish the transformation model between clothing,and extract the clothing deformation from a lot of affine transformation by the deformation characteristic extraction method above.Finally,To obtain the fitting clothing which have the same style with the sample dress,we transform the standard size clothing with the clothing deformation.To simulate the virtual fitting base on the data-driven method of fitting simulation,we select 20 kind of clothing samples to construct the clothing instance set of static A-pose virtual human,moving virtual human and the virtual human in some complex poses.Besides,we select 8 new types of clothing as the dressing clothing.3)A evaluation method of fitting effect is proposed.Our evaluation of the fitting is based on the fitting degree,the sense of reality and the degree of style change.The fitting degree and the sense of reality could judge by human eyes,and the degree of style change could be measured by the change of the discrete curvature of clothing.Experiments shows that our method has good effect on the pear human model,the apple human model and the kid model.