A study on compression of light-field imaging

Abstract

Although the concept of capturing light-field data has been around for decades, it has only gathered concrete interest in the last few years through the advent of digital processing and the introduction of the handheld plenoptic camera. Light-field imaging provides an extended gamut of features over traditional imagery, such as digital refocusing, slight perspective shift, and a variable depth of field, all of which are made available post-image capture. This could potentially see conventional photography becoming obsolete. With this added degree of freedom come large amounts of data which need to be compressed as efficiently as possible to allow transmission on bandwidth-limited channels or storage on capacity-limited devices. This study reviews, analyses, and benchmarks the current coding techniques available in the area, using appropriate measures including the Peak Signal-to-Noise Ratio (PSNR) and Structural SIMilarity (SSIM) index. The literature evaluation carried out and current efforts by the Joint Photographic Experts Group (JPEG) to provide a standard for light-field data, point to a solution that uses JPEG for compatibility with legacy systems. The standard JPEG compression scheme is modified and implemented, improved to be better suited for light-field image data. Thorough testing is carried out on the implemented solution, whereby a detailed evaluation is made through the comparison of the extracted results with the reviewed literature. Possible solutions for new frameworks are proposed, together with potential future work.