Real-time image restoration of sub-aperture images for ultra-thin light field cameras

Abstract

Ultra-thin light field camera is a light field camera made by attaching the lens array in front of the image sensor. The sub-aperture image quality is very low in ultra-thin light field camera due to the trade-off between spatial-angular resolution and non-existence of objective lens. Therefore, the restoration of sub-aperture image is necessary process to utilize the spatial information from the ultra-thin light field camera. In contrast to the previous works that requires complex computation or non-real-time, we propose a laplacian pyramid-based image fusion that can perform the enough restoration to improve the quality of the ultra-thin light field camera. We also present the ultra-thin light field camera having 24 sub-apertures, integrated with embedded computer so that can provide maximum portability. As a result, our ultra-thin light field camera system provides the restored image in real-time on embedded system. We also evaluate our restoration method with various criteria using our ultra-thin light field camera, our method shows better performance than other methods.