College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 200234, China.
Distribution Grid Dispatching and Control Center, State Grid Qingdao Power Supply Company, Qingdao, Shandong 266001, China.
Biomed Res Int. 2022 Jun 13;2022:8324438. doi: 10.1155/2022/8324438. eCollection 2022.
Different from traditional images, light field images record not only spatial information but also angle information. Due to the large volume of light field data brings great difficulties to storage and compression, light field compression technology has attracted much attention. The epipolar plane image (EPI) contains a lot of low rank information, which is suitable for recovering the complete EPI from a part of EPI. In this paper, a light field image coding framework based on EPI restoration neural network has been proposed. Compared with previous algorithms, the proposed algorithm further takes advantage of the inherent similarity in light field images, and the proposed framework has higher performance and robustness. Experimental results show that the proposed method has superior performance compared to the state-of-the-art both in quantitatively and qualitatively.
不同于传统图像，光场图像不仅记录了空间信息，还记录了角度信息。由于光场数据的体积庞大，给存储和压缩带来了很大的困难，因此光场压缩技术引起了广泛关注。对极面图像（EPI）包含大量低秩信息，适合从部分 EPI 中恢复完整的 EPI。本文提出了一种基于 EPI 恢复神经网络的光场图像编码框架。与之前的算法相比，该算法进一步利用了光场图像固有的相似性，具有更高的性能和鲁棒性。实验结果表明，与现有技术相比，该方法在定量和定性方面都具有优越的性能。
