Discarding invisible polygons

Hello. I’m sorry for the machine translator.
I have a question:
read and searched about discarding invisible polygons, and nowhere did I find normal maps to be applied. Well, it’s like an atlas of the starry sky, only instead of coordinates - the angles at which the polygon is rotated to the center of the object. At the first stage, this would help to discard half of the polygons from processing, meaning the angle of rotation of the camera. Please tell me - maybe I’m looking badly.

In general, I thought about creating my own engine, and I do not want to use known methods - I want to do something new. I understand this is difficult and very time consuming, but you can create some kind of baseIn general, I thought about creating my own engine, and I do not want to use known methods - I want to do something new. I understand this is difficult and very time consuming, but you can create some kind of base …
and I have some work on algorithms, I want to try to integrate them into a graphics engine.

someone opened the topic
https://forums.developer.nvidia.com/t/photoshop-absturz-bei-nvidia-normal-map/165041

but as I understand it is completely different

and here
https://dtf.ru/gamedev/118440-imitiruya-obem-kak-karty-normaley-pomogayut-zamenit-vysokopoligonalnye-obekty

https://habr.com/ru/post/481480/
but here the vertex normals

and to discard invisible polygons you need a polygon normal map

and that’s not it
http://www.malbred.com/3d-grafika-3d-redaktory/sovremennaya-terminologiya-3d-grafiki/normal-mapping.html

the image of my map is simplified, I thought everyone understood me, but on our forum it turned out that no one. in the evening I will definitely complete the image and description, the fact is that I work practically every day, and immediately could not make a quality one - I thought that everyone understood.
https://drive.google.com/file/d/1qjbdgELnAPHt7VlfO7wL6oyAkVCdYlHA/view?usp=sharing

I apologize, since one image is not enough to describe the essence of an easily understandable one, I make several images in 3D. Therefore, I do not have time today. But the method immediately discards half of the invisible polygons (according to the previously compiled polygon normal map), depending on the orientation of the camera. It will be ready tomorrow. I apologize.

Images on this site
https://sites.google.com/view/hummingbird-graph-engine

Images on this site.
I’m preparing a few more images and a description. Unfortunately, I can’t do it quickly, because I do it in my free time.

And it turns out the following, and then you need to think even more, but at the moment:
since this map is not responsible for the positions of objects, but only for their landmarks, then on this map as a camera we would not rotate around its own axis - this will not affect anything. Therefore, the supposed globe will have two axes, and will be devoid of medians, but have only parallels, since the medians are responsible not for the angle, but for the position. And the task is also that you need to consider how the capture area will look on a flat map.
.

And it is obvious that one axis should be tracked to the leading X to Z (turning Y does not care, since this map is not responsible for positions). That is, there must be a leading axis and a driven one, which will distort the transmitted data. Here you still need to think.
You need to think about how to work with the tracking of the second axis.

The site has been updated, tomorrow a page in English will appear, or, most likely, duplicate texts will appear.

Of course, this map does it for preprocessing polygons (grossly discard those polygons that are definitely not needed).

In fact, in order not to invert the normals for the map, it is enough to invert the direction of the camera for this processing.

And a very important point is that unnecessary polygons will not be discarded, but the necessary polygons will be selected from the map (discarding polygons is unnecessary operations)

Added images and description.
Added images and description.
Tomorrow I will make a text in English

I wanted to take a break for three days, but unfortunately I can’t stop, as it turned out that my reverse wave algorithm can be easily integrated into maps. In the evening I will update the site of the engine.

Although the reverse wave algorithm may not be mine, but I have not found such information anywhere a few years ago about a similar one, and still. Found out only about the direct wave algorithm.

the approach of the camera to the object can be considered as the approach of the cutting surface of the view in its direction to the edge of the sphere

Unfortunately, the break was a little longer. This happened due to the urgent need to resolve something on other matters. The engine has only been corrected for this image.

I haven’t done a translation yet, but there is not a lot of text there, I will definitely make a page in English.

Thank you very much for your patience for not pulling your hand. I made a page in English. I corrected the image, the next step is the formulas for sampling the covered area from the map.

From now on, everything will be much more thorough and detailed, and the old, which requires it, will be clarified.

The page is in English.

Thanks again for your patience. Everything is done in free time, and here you have to think more than work. But when the time comes to write the code, a lot will be predetermined.