Image processing: what is occlusion?

Question

I am developing an image processing project, and in many scientific articles I come across the word occlusion , what do occlusions mean in the context of image processing? The dictionary gives a general definition. Can someone describe them using an image as a context?

Answer 1

Occlusion means that there is something that you want to see, but cannot because of some property setting of your sensor or any event. Exactly how it manifests itself or how you deal with the problem will depend on the problem.

Some examples:

If you are developing a system in which it tracks objects (people, cars, ...), then occlusion occurs if the object you are tracking is hidden (closed) by another object. Like two people walking past each other, or a car that rides under a bridge. The problem in this case is what you do when the object disappears and reappears again.

If you use a range camera, then occlusion is an area in which you have no information. Some laser rangefinders operate by transmitting a laser beam to the surface under investigation, and then install a camera that identifies the point of impact of this laser in the resulting image. This gives the 3D coordinates of this point. However, since the camera and the laser are not necessarily aligned, there may be points on the test surface that the camera can see, but the laser cannot get (occlusion). The problem here is more related to the sensor setup.

The same thing can happen in a stereo image if there are parts of the scene that are visible only to one of the two cameras. From these points, it is obvious that range data cannot be collected.

There are probably more examples.

If you indicate your problem, then perhaps we can determine what occlusion is in this case and what problems it entails.

Answer 2

The problem of occlusion is one of the main reasons why computer vision is generally difficult. In particular, it is much more problematic in tracking objects . See the numbers below:

Please note that the lady's face is not completely visible in frames 0519 and 0835 in contrast to the face in frame 0005 .

---

And here is another picture where the man’s face is partially hidden in all three frames.

---

Pay attention to the following image how the tracking of a pair in the red and green bounding box is lost in the middle frame due to occlusion (i.e. partially hidden by another person in front of them), but is correctly tracked in the last frame when they become (almost) completely visible.

Photo courtesy of: Stanford, USC

Answer 3

Occlusion is one that blocks our gaze. In the image shown here, we can easily see people in the front row. But the second line is partially visible, and the third line is much less noticeable. Here we say that the second line is partially covered by the first line, and the third line is closed by the first and second lines. We can see such occlusions in class rooms (students sitting in rows), road junctions (vehicles waiting for a signal), forests (trees and plants), etc., when there are many objects.

Answer 4

Since the other answers explained the occlusion well, I will only add to this. In fact, there is a semantic gap between us and computers.

The computer actually sees each image as a sequence of values, usually in the range of 0-255, for each color in the RGB image. These values ​​are indexed in the form (rows, columns) for each point in the image. Thus, if the objects change their position relative to the camera where any aspect of the object is hidden (let the human hands not be shown), the computer will see other numbers (or edges, or any other functions), so this will change for the computer algorithm to detect, Recognize or track an object.