|
The perspectives of light-field images taken by lens-based light-field camera are fixed due to the fixed position and size of lenslet. Unlike lens-based light-field imaging, lensless light-field imaging technologies encode light-field information of scene through a diffuser, and then reconstruct the light-field images by computational algorithms. In this paper, a commercial holographic diffuser is employed in the lensless light-field imaging system, and a flexible complementary segmentation method of sub-aperture region is proposed for the Point Spread Function (PSF). The surface of the holographic diffuser is composed of relief microstructures, which are pseudo-random and non-periodic. Hence, any region of the holographic diffuser can encode and recover light information of object independently, just like sub-aperture imaging from corresponding perspective. Firstly, the PSF of lensless light-field imaging system is calibrated, and the encoded image of scene is also captured. After that, the PSF is divided into two complementary regions as sub-PSFs, corresponding to different sub-aperture regions. Then sub-aperture images of scene are reconstructed with corresponding sub-PSFs through reconstruction algorithm, achieving light-field image of selected perspective. Properly dividing the sub-PSF and reconstructing sub-aperture image, the sub-aperture image array of light-field can be obtained. With these light-field images, digital refocusing can be achieved finally. Different from existing segmentation method of PSF in lensless light-field imaging, the complementary segmentation method divides PSF into two sub-PSFs while these tow sub-PSFs are complementary to each other. This allows us to arbitrarily select the position and size of sub-PSF flexibly, namely flexible perspective selection. Meanwhile, decreasing the sub-aperture images to be reconstructed in algorithm can reduce error in reconstruction. Experiments show that, the complementary segmentation method can freely select the position and size of sub-aperture region while reducing the reconstruction error of lensless light-field imaging system significantly.
|
