Light field (LF) reconstruction is a technique for synthesizing views between LF images and various methods have been proposed to obtain high-quality LF reconstructed images. In this paper, we propose a progressive exploration network using efficient channel fusion for light field reconstruction (Prex-Net), which consists of three parts to quickly produce high-quality synthesized LF images. The initial feature extraction module uses 3D convolution to obtain deep correlations between multiple LF input images. In the channel fusion module, the extracted initial feature map passes through successive up- and down-fusion blocks and continuously searches for features required for LF reconstruction. The fusion block collects the pixels of channels by pixel shuffle and applies convolution to the collected pixels to fuse the information existing between channels. Finally, the LF restoration module synthesizes LF images with high angular resolution through simple convolution using the concatenated outputs of down-fusion blocks. The proposed Prex-Net synthesizes views between LF images faster than existing LF restoration methods and shows good results in the PSNR performance of the synthesized image.

In this paper, we propose a new light field (LF) reconstruction network that increases LF angular resolution. The proposed method consists of two parts that extracts the initial feature map and refines the extracted initial feature map. In order to efficiently extract features from input images, we rearrange the input images into a macro-pixel image using pixel shuffle and then extract the initial feature map using successive convolution layers. The refinement network continuously extracts inherent correlation information using a dense back-projection structure. Finally, we are able to stably train the network and reconstruct high-quality LF images by connecting the initial feature map and the output of the refinement network with long skip connection. Simulation results show that the proposed network outperforms other existing methods in terms of execution time and reconstructed LF image quality.

We propose an attention-based back-projection network that enhances light field reconstruction quality by modeling inter-view dependencies. The network uses pixel shuffle to efficiently extract initial features. Spatial attention focuses on important regions while capturing inter-view dependencies. Skip connections in the refinement network improve stability and reconstruction performance. In addition, channel attention within the projection blocks enhances structural representation across views. The proposed method reconstructs high-quality light field images not only in general scenes but also in complex scenes containing occlusions and reflections. The experimental results show that the proposed method outperforms existing approaches.