I read somewhere in SE about stereo and mono videos. Stereo videos are captured by using two cameras with which we are allowed to measure depth accurately. But what about still images? what is difference between stereo image and mono one?
Another question: if I apply an algorithm on a stereo data set, the same result will be acquired as mono data set? to what extent segmentation of stereo images is different in comparison with other type of images?
Have you ever wondered why people and many animals have two eyes? By using two eyes we can have a perception of depth. The principle is very simple, while you are looking the screen keep your head fix and close one of your eyes, try to watch all the things behind your screen, now open your closed eye and close the other one and try to watch the same things, can you see the difference?
Each eye send information to our brain. This information is almost the same, if you could have two pictures of your both eyes and you try to overlap this pictures you will see that there are big displacements in the nearest objects to your eyes while the displacements in the objects that are far away will be very small. These observed displacements are translated to the depth of the objects.
Now other experiment that could help you to understand this concept is this: Try to keep your view fix close your eyes for a while, now open only one eye but try to don't move your view, try to touch the things around you, but remember keep your view fix, is it simple?. Now with one of your eyes closed, move your head and try to do the same, is it easier?. By moving your head, you are giving information to your brain of how far are the objects because in the nearest objects the observable displacements will be bigger. Of course you should consider that the eyes have also other kind of cells that help us to perceive depth. But in a simple stereo pair of cameras the information that we could recuperate is only the distance of the same point in each image, the disparity map, and there are a lot of ways to recuperate it, optical flow is very common. This kind of information is normally used to know the 3D space and the segmentation is applied comparing the the distance between the cameras and the objects.
While by using only a camera you can compare only the intensities changes and try to detect the main changes by using a lot of metrics. Is very common to use this kind of systems in tracking of objects by keeping the camera fix and trying to omit the things that aren't moving in the scene.
In stereo video/images you have more information per frame/image allowing for creating a 3D presentation of the image/video signal (depth). You may create mono images/video from the 3D. You are not able to create a 3D presentation from a mono image/video signal. That the reason that old movies, that were not captured with two cameras from two slightly different direction are not convertible into 3D.
