Uncalibrated Synthetic Aperture for Defocus Control

Exaggerated defocus can not be created with an ordinary compact digital camera because of its tiny sensor size, so it is hard to take pictures that attract a viewer to the main subject. On the other hand, there are many methods for controlling focus and defocus of previously taken pictures. However, most of these methods require purpose-built equipment such as a camera array to take pictures. Therefore, we propose a method to create images focused at any depth with arbitrarily blurred background from the set of images taken by a handheld compact digital camera moved randomly. Using our method, it is possible to produce various aesthetic blurs by changing the size, shape or density of the blur kernel. In addition, we confirm the potential of our method through a subjective evaluation of blurred images created by our system.

Synthesis of smooth defocus from limited number of input images


If we use limited number of input images for refocusing (defocusig) by synthetic aperture , the background is very cluttering and ugly. Upper left picture shows 5 input images for our experiment, and the upper right picture is the result of synthetic aperture without interpolation.


Therefore, we generate a lot of picutures between given viewpoint using interpolation. The input image is devided by Delaunay triangulation using found feature points, and then warped to generate the interpolated picutres using the disparity among observed pictures.


Upper left picuture is a result to focus at the foreground objects, and the left is focused at the background. We generated 100 pictures by interpolation to generate such smooth and beautiful defocus.


This picture is just for comarison, left : input image, right : generated image

Estimation of viewpoint locations


We used simple technique to estimate the relative viewpoint locations using the average of disparity. Upper left figure shows the actual location of the 81 viewpoints for evaluation, and the right is estimated locations. NOTE THAT the baseline length is so subtle (span is 2.5mm each) while the distance to the object is around 1000mm, so we can say it is possible to estimate the location of viewpoints sufficient to defocus generation.


These pictures are generated from only THREE pictures taken by handheld digital camera, so the locations of the viewpoints are totally unknown. Focused at fouregound object (left) and background (right).


Performed with :

Natsumi Kusumoto, 2007.9-

See also :