Solving model fitting, cloth clothing mapping algorithm

Reference
Demo1 Http://bubaishi.com/bin/NetWeare.dll/LoginCheck
Demo2 http://www.soouya.com/#product/modellist?index=0&moteid=457

The above two examples are given a particular garment, which is replaced with a cloth to fill the tile in the specified area.
The implementation of the above is the back end of the completion of the picture synthesis.

I understand:
Three pictures:
First Model Complete P1
The second model fills the area diagram P2
The third one is the pattern to fill P3

The question is how to separate the P2 from the P1, and let the P3 be tiled and then synthesized into P1.
Do not know this P2 separate size and P1 is consistent, if inconsistent that is not the position is difficult to calculate AH.
We usually fill in the watermark is a position, but the clothing buckle area is irregular it.
Do you need special treatment in addition to the Pleated wrinkles section?

Please help analysis under, the best is a viable solution, thank you!

Reply content:

Reference
Demo1 Http://bubaishi.com/bin/NetWeare.dll/LoginCheck
Demo2 http://www.soouya.com/#product/modellist?index=0&moteid=457

The above two examples are given a particular garment, which is replaced with a cloth to fill the tile in the specified area.
The implementation of the above is the back end of the completion of the picture synthesis.

I understand:
Three pictures:
First Model Complete P1
The second model fills the area diagram P2
The third one is the pattern to fill P3

The question is how to separate the P2 from the P1, and let the P3 be tiled and then synthesized into P1.
Do not know this P2 separate size and P1 is consistent, if inconsistent that is not the position is difficult to calculate AH.
We usually fill in the watermark is a position, but the clothing buckle area is irregular it.
Do you need special treatment in addition to the Pleated wrinkles section?

Please help analysis under, the best is a viable solution, thank you!

According to my understanding, the current deduction of PNG, non-background part of the pixel padding is a problem.

Yes, it's interesting. yy ~

1. The whole model is going to be there.

2. Deduct the area information that can be replaced by the model's full picture setting

3. According to the light effect of the pictures of different models, the clothes folds, the recording record from a normal map to the whole transformation process script that conforms to the model photo effect, this step is the most difficult and core, it can replace the map-render (IMG) by providing API interface.

4. Captures the specified content from the image by the predefined map shape size of the selected model picture or uploaded picture

5. The image is passed to the Renderapi interface, and the final picture is applied to the transformation process.

6. Combine the rendering of each part into the final piece

