This week we talk about MotionVector,I think it’s a bit like a normal clean up, but the difference is that in this clean up we’re not dealing with a steady shot and a steady object that doesn’t move, but more of a marker point on the body that changes in light and shape。
Before we started working on the clean up we learned about a node called regrain, because when we made the clean up we may have thought we had done a good job on the final output, but in fact we could see in the other exaggerated colour gamut that we had overlooked some of the pixels. This node helps us to make our keyed clean up more realistic to the original shot.
We then started to use different ways of doing complex clean up. we could have tracked our roto manually, but this method is obviously not very clever and would only be considered if all other methods failed.
We then introduced the multiplication and division of merge and blur together to create the effect. Because the blur acts as an averaging of the surrounding colours, it makes the effect more realistic and natural.
But all of this works on the premise that we need to manually adjust the roto, or use a single track. so one logically simpler way to understand this is to use tracking points and panning, which ensures that the roto comes out in the same position as the face. I think this method achieves good results with constant lighting and the structure is very easy to understand.
Similarly we can give each point a trace and then rotopaint it separately.
There is also the reverse thinking, using blur and edge blur to reverse fill the mask after making the trace can also work well.
Next we talked about a very important part called the uv map, which I think is closer to the dynamic stmap, which gives information about movement. So we can use this node to clean up the face and add textures. It is also possible to get motion blur. However, it is important to note that it has certain requirements for the shot and does not fit all shots perfectly.