In my algorithm I shift all the images by w*shift amount where the shifts are as follows is:
xshit = index/17 and yshift = index%17 where index is for the list if images i have which go from 0 to 289
This works because the images are taken in a lightfiled format through a 2d grid perpendicular to the optical axis.
I display the results by focusing for different w values
|
||
|
|
|
|
|
|
On both sides (+ and -) we can see that increasing the w component changes which band of image gets focused and for 3 and -3 the focus band is outside the image.
For this part the idea I used was that avergaing over a small subset of my lightfield images gives a narrow aperture while averaging over a bigger set imitates a wider aperture.
I set my radius to be the image at the index location 162 which is located at Toys[9][9] on the grid.
Then I do my aperture adjustments in radii aroud this grid.
Below are the image results for different apertures for the dataset:
|
|
|
|
|
|
|
|
|
I learned through this project that light camera is a very powerful technique to implement different perspectives on your images. But at the same time it is very easy to implement. Very cool and fun!!