Low-Complexity Compression of 360° Still Images

This paper presents a novel approach for approximated 360° still image compression that combines latitudeadaptive quantization with low-complexity transform and quantization. While addressing the redundancies inherent in the equirectangular projection format, the proposed method is multiplication-free, requiring only addition and bit-shift operations. We experiment with different low-complexity transforms, showcasing the plug-and-play nature of our approach. We test our methods in a popular benchmark – including 4K, 6K, and 8K images – and compare them against their exact counterparts. We show that our approximated compression methods lead to no higher than 4–7 dB quality loss at small bitrates, providing efficient alternatives for resource-constrained devices.