INTERFERE: a high-throughput, view-selective codec for holography

Abstract

We showcase INTERFERE, a hologram compression framework selected as the basis of the JPEG Pleno Holography standard. It supports view-dependent coding with simultaneous spatial and angular random access. INTERFERE utilizes an adaptive quantization mechanism that can assign a variable bit width across small phase-space regions. This ensures that the transform coefficients are compactly represented before entropy coding. In this work, we design a new entropy coding mechanism with division-free binary arithmetic coding, allowing us to better capitalize on the compact quantized representation. We also create new probability models for driving the binary arithmetic coder, which can fully harness SIMD instructions while exhibiting significantly smaller memory requirements, enabling it to reside in the CPU cache. Speed-ups ranging from 6x 600x in decoding and encoding times on the CPU were achieved over our previous solution without compromising rate-distortion performance. With this work, we propose a practical hologram codec that is class-leading in rate-distortion performance, ease of random access, and encoding/decoding throughput, thereby providing a practical solution to one of digital holography’s most pressing challenges.