Is it possible for the engineers to take say 100 clear pictues and deliberately take 100 blurry pictures of the same pose and train/program the software to transform the pixels from blurry to clear?Can engineers make a super photo image enhancer by programming many clear and blurry pictures of the same shot?
All that would do is teach the computer how to enhance that image. The algorithms produced would not work (or not work as well) on another image.
But essentially this is one of the ways image enhancement techniques work. By learning the conventions of shapes, lines, shadow, light and substituting this knowledge for the missing information. The thing is in the end it is not an 'enhancement' it is a supposition for what the clear image should be like based on what is known.Can engineers make a super photo image enhancer by programming many clear and blurry pictures of the same shot?
The answer to your question is: No . . . this can't be done even though it seems like it should be possible. The problem is that when an image is blurred for any reason (improper focus, camera shake, subject movement) data is lost. The proper location of edges are lost, contrast and color are lost. This information cannot be recovered from the information that is left.
This is different from correcting distortion in an image. For example, you can correct the distortion of a fish eye lens because you only need to move lines and shapes. All of the information remains in the fish eye photo. None needs to be recreated.
In fact they do something like this - especially when they are trying to reconstruct blurry images from satellites, and even nuclear particle colliders. But in general it is NEVER good as just taking the picture correctly. This kind of image processing is reserved for situation with (a) lots of computing processor power (2) absolutely no other recourse.
Good luck...
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment