Yes, you are right about scanners, there are really not cheap if you want good results and we all like the super scalability and cheapness of photogrammetry here ! you can really push the limits with “reasonable” investments and training.
I think it can definitely work, there is some detail on the statue that can help.
I would not work with a turntable - my experience with that is that when you have an object that is a bit limit (too shiny, not enough features), this is worst when using the turntable.
You need to put a lot of fully featured items all around the statue : use newspapers, magazines, (just be sure they don’t move)… this will help alignments.
I would not work above f/16. It depends on your lens of course, but at f/16 and f/22 you have a big loss of quality (diffraction) - especially if you are at 100 iso on a tripod… You have to do some tests to find the best tradeoff between depth of field and sharpness - some where between f/5.6 and f/11.
Work at a fixed white balance (measure it on a chart if you want a good accuracy)
And, very important, take a lot of photos with great overlap. All around, then closeup details.
Don’t forget to include a scale or a visual with known markers, where you can put a reference distance.
The pattern of chromatic noise consist of :
-generate a picture of chromatic noise in GIMP : you will end up with a pattern of unique dots of random colors. You can experiment with several settings for resolution, density of noise, size and upscaling (I usually generate a 960x540 pattern that I upscale in nearest mode to a 1920x1080 to have 2x2 colorfull tiles)
-use a video projector to project on the piece. Find the best position to have a sharp projection on the piece.
-You will maybe need several projectors (in that case, you need to generate other patterns, do not reuse a pattern).
-take pictures all around with and without the pattern videoprojection.
-align pictures.
-use the pictures without the pattern to texture.
The pattern of chromatic noise really helps RC to understand the surface (because this is a lot of unique features) and achieve much better accuracy. Maybe I will write some tutorial if more people are interested in the technique.