From the perspective of game developers and early photography.

Since I stepped into the photography industry and met some friends engaged in the art industry, I found that there is a problem worth thinking about: their understanding of "perspective" is completely different from our 3d industry.

For practitioners in the 3d industry, the MVP transformation in the rendering pipeline is more or less known, so most people will take it for granted that perspective is only related to the degree of perspective transformation.

(The actual position of the object remains unchanged. Just zoom in along the positive Z axis of the camera, so that the first square will always be in the same position in the picture. )

Practitioners in the art industry/photography industry told me that the distortion of perspective is only related to the distance between the subject and the camera. For example, in the following two cases, you will find that although different focal lengths are used, the second one can be superimposed with the first one by cutting and magnifying.

(Don't change the position and angle of the camera, and don't change the actual position of the object)

In fact, I think there is nothing wrong with these two views. It's just that the way of expression is different from their respective applicable occasions, which leads to the deviation of understanding between the two sides. Moreover, people on the Internet have been arguing about the right and wrong of these two understandings, so please let me sort it out from my own point of view.

In the objective world, the farther an object comes to human eyes, the smaller it is. Therefore, we understand that the concept of perspective can actually be simplified to the near big and far small.

But how to explain the perspective? We look at it from two angles.

In order to make things clear, here is a little unprofessional language to explain how perspective is realized in three-dimensional software.

In the world space, all objects have no perspective effect at first, which is often called orthogonal perspective. In order to achieve the perspective effect, we can create a trapezoidal space similar to a pyramid, wrap the object to be rendered in it, and then restore this trapezoidal space to a cube/cuboid, so that the wrapped object will also deform with the change of space shape. In this way, we can make these objects of the same size produce a perspective effect, and then we just need to render them normally with the camera.

(Suppose the ball is a camera and the frame is a packaging space.)

In principle, it is not difficult to see that the gradient of trapezoidal space can be changed. The more oblique, the stronger the perspective effect, and the more deformed and bulky the object closer to the camera. On the other hand, the closer an object is to a shape without deformation, the smaller its volume will be.

However, as users of engines or 3D software, we will not manually interfere with the slope of this trapezoidal space, but use simple and understandable parameters such as fov value and focal length to adjust its final slope.

In modern times, most cutting-edge 3d software pays attention to the physical foundation, and the algorithm may become more complicated than the pyramid mentioned above, so it is common to control the virtual rendering camera to approach the real camera effect with "focal length" in professional renderers. (The only difference is that the virtual camera has no lens distortion, and the final rendered picture will not produce optical distortion, but will only produce perspective stretching and extrusion deformation because of different focal lengths. )

So for 3d practitioners, "perspective" refers to the size contrast between the object closest to the camera (the largest) and the object farthest from the camera (the smallest) in a limited distance space, which is caused by perspective transformation. The greater the contrast, the greater the perspective at this time, and vice versa.

For photographers, it is impossible for us to show the audience comic photos so that they can feel the information in the photos. First, it is stupid to do so, and second, there is not enough spare sd card for us to shutter every scene.

Therefore, rather than taking dozens of photos in a single action, photographers are more willing to work hard on the photos they want to take (one frame), choose the appropriate focal length, determine the scene to be used and the shooting angle, and then press the shutter. As long as you grasp the movements according to your own inner standards, the later film selection and secondary composition will be much more convenient and worry-free.

but ...

If you take the wrong lens on the day of shooting, take 2470 to 1635. It is well known that telephoto lens will compress objects with different distances in the focal plane direction, while short focus will expand in the focal plane direction. What should we do? We are afraid of being killed by customers on the spot with a wide-angle short lens like 1635.

At this time, the intuitive factor that affects perspective-distance comes in handy. The pixels of modern cameras are generally not low, not to mention the 5d4 in our hands. 16mm and 50mm cubes, didn't we say that before? Except that the focal length of the camera has changed, the position of the camera and the position of the object in the world coordinate system have not changed. The two cubes photographed at this time are actually exactly the same except for the size in the picture.

Depending on this principle, we can also use 35mm hold 1635 to shoot a face with 70mm effect. We just need to stay a little away from the subject and stare at the other person's face in the viewfinder until the deformation of the face doesn't look too big, so we can press the shutter. As long as the shooting is clear enough and the noise and exposure are properly controlled, the later editing can still get a bust with normal perspective and save the film.

But I thought for a moment, in this case, I might as well try to find a chance to change the lens into a small spittoon or shoot 85 hard on the spot. If it really doesn't work, convince customers to give a bust and a close-up ... Quan Huafu's pixel is forced to be cut into close-ups, even if it is a bit overwhelming, unless we have a large frame in our hand.

Therefore, for the practitioners in the photography industry, "perspective" usually means that the cameraman uses a fixed focal length in a single picture, and by changing the distance between the camera he holds and the subject, the subject in the picture is deformed in different degrees at different distances, and according to this method, more available perspective deformation schemes are created on a single photo that cannot be dynamically refreshed.

I believe that after reading the above two perspectives, you will still be a little confused. Here is a summary.

In my opinion, both "focus changes perspective" and "distance changes perspective" are the laws discovered by predecessors when exploring the objective world, so they are all correct and available, only used in different occasions.

For game practitioners, if these games refresh 60 frames per second, and their positions and perspectives change all the time, they need to be as close as possible to the visual habits of human eyes (human eyes are a superimposed 16:9 screen, and the focal length is equivalent to 50mmf8 lens, so they should be as close to this focal length as possible, so that players can not easily get carsick when playing with mobile phone cameras, which is also commonly known as dizziness. Therefore, the system of changing the viewing angle by changing the focal length /fov value is applicable in this case.

For photographers, every photo that records a wonderful moment is precious and cannot be copied. Unlike videos and games, they won't happen again as long as they are missed, so saving time is the most important thing. Documentary photographers, for example, need to race against time to capture the once-in-a-lifetime moment of customers, and it is even more impossible to waste time changing lenses (generally a 2470 is used from beginning to end). The focal length of Wanyishou can't reach the compression effect of telephoto, and the photos of that feeling have to be taken, so the very intuitive and easy-to-use system of changing the far and near perspective has become a unique rescue means at this time, and it has also left more possibilities for the subsequent secondary composition.

So the significance of focal segment is to determine the maximum viewing angle in the picture, because you can't get the distance segment outside the picture by interception.

The law is dead and man is alive. Making good use of these laws discovered and summarized by predecessors is the best understanding of their existence.