Graph Fourier Transform for Light Field Compression

This work proposes the use of graph Fourier transform (GFT) for light field data compression. GFT is a tool developed for the emerging field of digital signal processing on graphs, which combines graph theory and classical DSP in order to exploit signal-related information present on graph structures. The proposed method explores the high correlation of residual images from light field. Simulations with real light field data indicate significant reduction in the number of coefficients for data representation; for instance, an 8.97% reduction was achieved while keeping smaller mean squared error when compared to discrete cosine transform-based compression.