Can we see a screenshots of your setup? maybe a few? make sure the whole Ae interface is visible.
My frozen frame is also masked off.
I thought I solved the issue by changing the quality and sampling box, but there are still issues, including some slight deformation on occasional frames.
Are you sampling colors with an expression?
Slight deformations and flickering sampled colors are common on thin lines and sampled colors when you are talking about moving and especially slow moving layers.
I don't see the modified properties of the layers giving you problems so I'm not sure what's going on. We need more details. All I know for sure is the null is animated, Layer 2 and 3 are parented to the null and layer 5 has a curve effect applied and you are using time remapping to freeze a frame at 8 frames. Please let me know if I missed something.
I'm not sampling any colours with expressions. I used the curves initially to match the gamma, then realised it was a broader issue. Basically, I have a patch of image, masked off from a clean frame of my original footage. I am applying this patch to the original footage via a null driven by transform data tracked in Mocha. I have two other patches/masked off freeze frames of clean plate, which display no issues in gamma change or distortion, they are driven by the exact same null's tracking data.
I have performed this technique literally dozens of times in this project and not had any issue, until this particular patch, it is the part of the picture where you seen the cctv camera and lamp post on the right hand side. If you look at the 2nd pic, you can see where the shift occurs, i hope.
I'm not following you exactly. I took both your screenshots and lined them up in Photoshop with the blend mode set to difference on the top layer to find the different pixels. This is what I got:
It looks to me like the upper right corner where the CTV camera and the light pole appear to be identical. It's the center of the frame that is different.
Compared to that:
Look pretty identical except for the left and lower part.
That said I'm assuming that the camera is moving so you tracked something in the shot and added that track info to the null. If the track isn't perfect and the tracking area is not very close to the replacement layers 2 and 3 that are attached to the null then lens distortion and tiny tracking errors could easily cause overlay issues with those shots. From my difference mode tests it looks like one of the layers is on the upper right extending down to nearly the bottom of the frame and the other replacement layer is just below and just right of center screen. That area is not nearly as perfect a pixel to pixel match as the upper right area. The rest of the image shows some definite differences between almost all of the pixels. Neither of these replacement layers appear to have soft edges, which is the first thing I would do to try and hide the problem. It also looks like you are not using any blend modes or other tricks to fix your composite.
So, the most likely cause, from what I see is either a slight flickering or shifting of the gamma in the main plate - the original footage - that could be caused by the street lights, and that flickering is making you think the freeze frames are flickering because of the behavior of the edges.
The other thing that may be going on, but it does not explain the apparent flickering in the upper right corner you are describing, is that there is a slight bit of imperfection in the tracking and that there is a bit of lens distortion at the edge of the frame. This can also cause problems with hard edges and blend modes. It's hard to tell because I don't know how much movement is in the shot.
I hope this helps. I hate these kinds of problems. They are very hard to diagnose and fix.
Hi Rick, Thank you for taking such an indepth approach and replication. I have taken further snippets with my gamma ramped up a little. The first pic is the first frame, the second pic is the second frame, and the third, I noticed when "gamma slamming" that this masked freeze frame patch also demonstrates the same gamma shift issue. I have not feathered the mask as it will clearly not help (at this stage) with such a gamma shift/blur occuring. You should be able to see that the gamma in the main plate from frame 1 to 2 does not show this discrepancy.
This is with quality and sampling set to best (biLinear) , if I choose draft I get tearing and stretching in certain frames of my still image however the gamma does not shift as can be seen above. Bi Cubic is the same as biLinear in this case. So, why is biLinear altering the gamma of my freeze frame from one frame to the next? Perhaps, it is an issue of some kind of blurring/denoising during the transformation of the image with biLinear, and why is Draft fine with handling it, but demonstrates occasional tearing/distorts? I have no idea and it is beyond my expertise.
Also, here is another shot from the same project, here there are no issues with gamma shift using BiLinear sampling.
First pic is of my freeze frame clean plates, there are around 8 of them;
Here, is the composite of them, notice there is no gamma shift, and it is this way throught the shot which is a long zoom out from this building;
So, it is only with my other comp, with deeper blacks that this is occuring.
It's actually more like it is softening the image, lowering its resolution drastically.
I had some help from Robert Kjettrup, who suggested that the problem was due to the complexity of the noise in the particular image I used. In maintaining this look, he suggested covering up the lack of detail with a separate (non-transformed) layer that was re-grained, then use a track mask in the areas you need to re-grain with this layer.
I normally denoise my images, but in this case I had left the shot as it was, so I was not used to encountering a problem such as this. Thanks also to Rick Gerard for his help too.
Forgot to mention, I first needed to denoise the footage before regraining it using match grain.