We present a new concept of scene adaptive imaging scheme for integral photography (IP), which is named
as "adaptive IP (AIP) imaging." Our proposal is to use variable focus lenses to compose the lens array for IP
imaging. Our scheme will greatly enhance the potential of free-viewpoint image synthesis from IP images, because
the sampling pattern of light-field can be optimized for the scene structure. We first introduce a theoretical model
describing how to optimize the light field sampling for the target scene, by using our virtual camera model in
the Hough transform space. We then describe our prototype implementation with 64 liquid lenses compactly
arranged in an 8 by 8 matrix, and preliminary results with it. Our imaging scheme can be regarded as an example
of Programmable Imaging, and will contribute to this new trend of imaging methods.