|
|
|
The idea was to mess about with image light field data sets to show focusing and aperture adjustments but using only the image data.
Given that many images have been taken at precise angles, we choose a center image and align other images accordingly to their offsets, based on their (u,v) coordinates, to that central images. Upon averaging the shifted images (shifting controlled by a scalar value alpha) we optain depth focusing.
|
|
|
|
Similar to the first part, this time we take the images in a certain range for aperture adjustment. This time we limit the images based on a new range offsetting from the (u,v) central one, and you will see a smaller aperture from the beginning showing its aperture focus is now only on an upper portion.
|
|
Overall I found this project to be quite an interesting one. Prior to this class, I did not know manipulation of image data can help recreate some camera functionalities. Light fields really change the game regarding manipulation
It is difficult to get working because the use of our own custom images are a lot harder regarding alignment because the light rays incoming are vastly translated even if a human hand would do a slight positional adjustment. Thus, it is not as easy without extremely precise adjustment (via machine holding a camera for example).