Hello . I apologize for errors in advance because I do not know English and I write through an interpreter.
When using RC, I encountered the problem of relative blurring of the texture when mixing images. I read suggestions on the new projection of textures, and the choice of sharp images for the priority of the view is certainly good, but as far as I know (maybe I'm mistaken) that algorithm is made on artificial neural networks, and things like neural networks are not always available.
What if we take the projection of each camera and bake it in a separate layer and add the tools for editing them ( brush erasing (stitches
and problem areas), layer mapping in%, and the function of mapping the layer on top of the rest in%)
It is clear that if in a scene of 2000 photos, then this option does not work (and for these tasks you need to keep and the old version of projection), but for scenes with objects in 100 -200 photos this is quite appropriate and after 1-2 hours of such editing it will be possible to get a texture with the maximum sharpness and without small stitches like in *** and in fact the quality of the texture will depend on the user and not on the software.
I think that simultaneously display 200 layers, too, do not work on this by default, for example, display 15-20 and then the user will choose the number of display layers by the characteristics of his computer.
Well, provide layers and tools in some convenient way.
I checked this method in 3d Сoat. I took a scene with 70 photos (stone) selected about 20 cameras in the manual and baked each in a separate layer / exported / then imported each layer to 3d Сoat and there already edited the layers. the texture turned out excellent. But 70% of the time I spent on baking each layer manuallyr and export / import to a third-party editor and I edited the layers in half an hour.
It's just a sentence and maybe you have other ideas on this, but in my opinion it's easier to implement than the super algorithm for projection.
Please sign in to leave a comment.