Distributed Compression for Large Camera Arrays

We address the problem of compression for large camera arrays, and propose a distributed solution based on Wyner-Ziv coding. The proposed scheme allows independent encoding of each view with low-complexity cameras, and performs centralized decoding with side information from additional views. Experimental results are given for two light field data sets. The performance of the proposed scheme is compared with independently coding each view using JPEG2000 and a shape-adaptive JPEG-like coder. The Wyner-Ziv coder yields superior compression performance at low bit-rates. In addition, there is a great reduction in encoder complexity when compared to JPEG2000.