The locked-down camera helps but there seems to be lots of folks moving/walking about. My thoughts are to roto each person individually.
Since the floor will be kept for the final composite, its entire area can be easily masked off to allow you to concentrate on head+shoulder masks for each individual rotoshape. As easy as the floor may sound, it does provide an issue. The floor exhibits different lighting scenarios, you'll have to decide how best to tackle this. Ideally, the final composite includes a darkened floor. From the images provided, you may be able to generate a luminance matte to provide you with a decent start.
Good on you for providing the images but it would have also been useful if you had provided more info on what the final composite is supposed to look like.
If the people don't turn their heads and shoulders much (eg. torso/head silhouettes get smaller but more or less maintain overall shape as people walk to train) You should be able to roto each individual very fast in Mocha if you have it.
Make sure you roto the people before the train since you can use the people rotos to mask out those areas when tracking the train.
3D animator and VFX specialist
regarding the final compo, the idea is to isolate people and train so we can compose a matte painting on the background...(simulating the other side of the train station).
at the moment what I was trying to make the silhouetted (as the floor don't move) and I was getting from the waist to the head more or less. the problem that I see is that as the people are crossing between them it is very hard to keep a shape for more than 6-7 frames... and if you change shapes very often then it looks jerky....
I gonna try the luminance matte just in case....(I tried before difference kind of mattes but none give me a decent result)
Hi Raul, it looks like a lot of work unless you can generate a usable difference or luminance matte. Another option is to darken the bottom part of the foreground floor to preclude roto-ing anything in the bottom 30% of the image - essentially where the leg action (inluding shadows) take place.
For this, you may want to create a single frame to obtain approval from the director/producer.
If you're stuck with roto then roto each person individual with separate masks for each pertinent body part. For example, the left leg should have at least three part - (1) knee and above (2) ankle up to knee and (3) shoe.
As you can see, it's a lot of work just for the legs and then the shadows require separate and similar treatment.
If I were to direct this shot, I would have shot each individual separately (with a chroma backdrop), relight in post and then composite them. Then shoot the train separately and then perform a final composite and apply final color grading touches.
I agree with you...they should had shotted this with a backdrop....but...
about what you were suggesting....I don't need to roto the whole leg and shadows, as I need to keep the floor...I will add an image so you can see it...
I kind of gave up the rotoscoping approach... the main problem that I'm finding is when the people cross ways between them, when that happens the shape change considerably and i get kerky animation....
other of the issues that I'm seeing is that even with the rotoscope that heads don't look "very accurate"
I keep trying to apply keyer to see if there is any work around, but the similarities of color between the heads and the people it's complicating everything
People crossing past each other from the cameras point of view should not present a major problem unless you’re using one roto mask for more than one head or shoulders at a time.
Just as Roland suggested, you must break up each person to be rotoscoped into its simplest shapes.
My approach would be to go into Mocha and boost the gamma on your footage to see if there is some surface details on heads and backs of people.
If there is then start by choosing one head closest to camera that is never occluded by anything and create a very simple circular X-Spline (name it HeadA-track for example) WITHIN the boundary of the head at roughly half way along the time line. Track forward and back. If it tracked well, turn off tracking for that layer. Now go back to half way on time line and create another X-Spline for the same head (name it HeadA-roto for example) and this time roto the head carefully for that frame. Turn off tracking for this layer too and link it to HeadA-track. You might need to make some minor tweaks to HeadA-roto in “AdjustTrack” to fix drift.
Now continue the process for all heads and shoulders separately. Remember layer order in Mocha is vital whilst tracking since occluding shapes already tracked need to sit at the top of the layer stack. Work from objects closest to camera first ending with the train last.
If a head crosses in front of someone you might need to help mocha a bit during the tracking but once you get into the swing of it, it shouldn’t be too painful!
3D animator and VFX specialist
Not having the moving shot in front of me I can not say that this will work for sure, but here's a crazy thought:
Try to see if you can create a animated clean plate- clean a still of the back wall from the first frame (including the train with the shadow and lights), and animate new shadow and lights coming from the train as it pulls in using the clip as reference. Match the grain and you should be able to get a clean plate that will allow you to pull a difference key.