Stereo Image Compression

Stereo image compression aims to efficiently represent pairs of images capturing the same scene from slightly different viewpoints, minimizing storage and transmission costs while preserving visual quality. Recent research focuses on developing neural network-based codecs that leverage bidirectional processing and cross-view information, often employing architectures like transformers and convolutional neural networks with attention mechanisms to exploit inter-image dependencies. These advancements improve both rate-distortion performance and encoding/decoding speed, impacting applications such as virtual reality, autonomous driving, and 3D imaging by enabling more efficient handling of multi-view data.