Geometry refinement for light field compression

Light field (LF) imaging has gained significant attention due to its recent success in microscopy, 3-dimensional (3D) displaying and rendering, augmented and virtual reality usage. Postprocessing of LF enables us to extract more information from a scene compared to traditional cameras. However, the use of LF is still a research novelty because of the current limitations in capturing high-resolution LF in all of its four dimensions. While researchers are actively improving methods of capturing high-resolution LF's, using simulation, it is possible to explore a high-quality captured LF's properties. The immediate concerns following the LF capture are its storage and processing time. A rich LF occupies a large chunk of memory ---order of multiple gigabytes per LF---. Also, most feature extraction techniques associated with LF postprocessing involve multi-dimensional integration that requires access to the whole LF and is usually time-consuming. Recent advancements in computer processing units made it possible to simulate realistic images using physical-based rendering software. In this work, at first, a transformation function is proposed for building a camera array (CA) to capture the same portion of LF from a scene that a standard plenoptic camera (SPC) can acquire. Using this transformation, LF simulation with similar properties as a plenoptic camera will become trivial in any rendering software. Artificial intelligence (AI) and machine learning (ML) algorithms ---when deployed on the new generation of GPUs--- are faster than ever. It is possible to generate and train large networks with millions of trainable parameters to learn very complex features. Here, residual convolutional neural network (RCNN) structures are employed to build complex networks for compression and feature extraction from an LF. By combining state-of-the-art image compression and RCNN, I have created a compression pipeline. The proposed pipeline's bit per pixel (bpp) ratio is 0.0047 on average. I show that with a 1% compression time cost and 18x speedup for decompression, our methods reconstructed LFs have better structural similarity index metric (SSIM) and comparable peak signal-to-noise ratio (PSNR) compared to the state-of-the-art video compression techniques used to compress LFs. In the end, using RCNN, I created a network called RefNet, for extracting a group of 16 refocused images from a raw LF. The training parameters of the 16 LFs are set to (\alpha=0.125, 0.250, 0.375, ..., 2.0) for training. I show that RefNet is 134x faster than the state-of-the-art refocusing technique. The RefNet is also superior in color prediction compared to the state-of-the-art

We propose a novel method of 5-D dynamic light field compression using multi-focus images. A light field enables us to observe its scene from various viewpoints. However, it generally consists of 4-D enormous data, that are not suitable for storing or transmitting without effective compression. 4-D light fields are very redundant because they essentially include just 3-D scene information. Actually, a method of reconstructing a light field directly from 3-D information composed of multi-focus images without any scene estimation is successfully derived, though robust 3-D scene estimation such as depth recovery from light fields is not so easy. Previously, based on the method, we proposed novel light field compression via multi-focus images as effective representation of 3-D scenes. In this paper, we extend this method to compression of 5-D light fields composed of multiview videos including the time domain. It achieves significant improvement in compression efficiency by utilizing multiple redundancy of 5-D light fields. We show experimental results using synthetic videos. Quality of reconstructed light fields is evaluated by PSNR and SSIM for analyzing characteristics of its performance. They reveal that our method is much superior to light field compression using HEVC at practical lower bit-rates.

Learning from EPI-Volume-Stack for Light Field image angular super-resolution

As light field images continue to increase in use and application, it becomes necessary to adapt existing image processing methods to this unique form of photography. In this paper we explore methods for applying neural style transfer to light field images. Feed-forward style transfer networks provide fast, high-quality results for monocular images, but no such networks exist for full light field images. Because of the size of these images, current light field data sets are small and are insufficient for training purely feed-forward style-transfer networks from scratch. Thus, it is necessary to adapt existing monocular style transfer networks in a way that allows for the stylization of each view of the light field while maintaining visual consistencies between views. To do this, we first generate disparity maps for each view given a single depth image for the light field. Then in a fashion similar to neural stylization of stereo images, we use disparity maps to enforce a consistency loss between views and to warp feature maps during the feed forward stylization. Unlike previous work, however, light fields have too many views to train a purely feed-forward network that can stylize the entire light field with angular consistency. Instead, the proposed method uses an iterative optimization for each view of a single light field image that backpropagates the consistency loss through the network. Thus, the network architecture allows for the incorporation of pre-trained fast monocular stylization network while avoiding the need for a large light field training set.

Light Field, a computational photography technique has various exciting features like refocusing and novel view synthesis. Previous works in light field photography have focused their work on creating light field images having a limited field of view. This limitation of light field images motivates the research for a acquisition system which can capture light field images with larger field of view. Recent works have been done to acquire light fields using a cylindrical arrangement of cameras which resolve the conventional complexity issue of acquisition of light field and subsequently merging them by essentially combining the capture and merging processes into a single exercise. We use this cylindrical system to generate a novel True Zoomed Light field panorama from the captured light field, which gives a wide 360? field of view similar to traditional panoramas along with the zoom applied to the whole panorama. The captured panorama can also be manipulated in various ways as is possible with other light field recordings. Also various zooming options which can be easily found in traditional camera, either tends to change the relative shape of the objects in image or introduces pix-elation effect. We use true zoom so as to improvise the zooming effect. We generate novel “True Zoomed Light Field Panorama” using our system. This type of panorama gives user a feeling of being closer to or farther from the whole environment at the same time.

Light fields capture 3D scene information by recording light rays emitted from a scene at various orientations. They offer a more immersive perception, compared with classic 2D images, but at the cost of huge data volumes. In this paper, we design a compact neural network representation for the light field compression task. In the same vein as the deep image prior, the neural network takes randomly initialized noise as input and is trained in a supervised manner in order to best reconstruct the target light field Sub-Aperture Images (SAIs). The network is composed of two types of complementary kernels: descriptive kernels (descriptors) that store scene description information learned during training, and modulatory kernels (modulators) that control the rendering of different SAIs from the queried perspectives. To further enhance compactness of the network meanwhile retain high quality of the decoded light field, we propose modulator allocation and apply kernel tensor decomposition techniques, followed by non-uniform quantization and lossless entropy coding. Extensive experiments demonstrate that our method outperforms other state-of-the-art (SOTA) methods by a significant margin in the light field compression task. Moreover, after adapting descriptors, the modulators learned from one light field can be transferred to new light fields for rendering dense views, showing the potential of the solution for view synthesis.

The light field image, also known as the plenoptic image, contains the information about not only the intensity of light in a scene but also the direction of the light rays in space. Since the light field image contains very rich photometric and geometric information, it will have very widespread application in the future. For example, immersive content capture for virtual and mixed reality presentation or depth from light field for auto driving applications. To be more specific, a light field image can be enhanced with physical models for an autonomous decision making process, which is also an important task of the Dynamic Data Driven Applications Systems (DDDAS) [11]. Besides, the rich geometry and photometric information contained in the light field can be updated with real-time measurements, which is a focus of DDDAS such as smart city related image and video processing tasks. However, to make the light field images easier to be utilized, one of the most important tasks is to compress the light field images efficiently so they can be easily distributed over the current communication infrastructure.

Currently, a raw light field image can be captured by a light field camera and decomposed into multi-views with different perceptual quality. A novel light field compression approach for a light field image is proposed based on this observation, where the raw light field image is decomposed into multi-views as pseudo-sequencing frames and reordered according to their perceptual quality. The resulting pseudo-temporal sequence is then compressed by high-efficiency video coding (HEVC) encoder. Experimental results show that the proposed method can achieve a better coding performance (up to 28% Bjontegaard delta bit rate reduction (BDBR) savings) compared with state-of-the-art reordering techniques using HEVC.

Light field images can record multiple information about the light rays in a scene and provide multiple views from a single image, offering a new data source for 3D reconstruction. However, ensuring the quality of light field images themselves is challenging, and distorted image inputs may lead to poor reconstruction results. Accurate light field image quality assessment can pre-judge the quality of light field images used as input for 3D reconstruction, providing a reference for the reconstruction results before the reconstruction work, significantly improving the efficiency of 3D reconstruction based on light field images. In this paper, we propose an Adaptive Vision Transformer-based light-field image-quality assessment model (AViT-LFIQA). The model adopts a multi-view sub-aperture image sequence input method, greatly reducing the number of input images while retaining as much information as possible from the original light field image, alleviating the training pressure on the neural network. Furthermore, we design an adaptive learnable attention layer based on ViT, which addresses the lack of inductive bias in ViT by using adaptive diagonal masking and a learnable temperature coefficient strategy, making the model more suitable for training on small datasets of light field images. Experimental results demonstrate that the proposed model is effective for various types of distortions and shows superior performance in light-field image-quality assessment.

Light field imaging is becoming a key technology, which provides users with a realistic visual experience through the capability of dynamic viewpoint shifting. This ability comes at the cost of capturing huge amounts of information, leaving the problem of compression and transmission a challenge. The encoder complexity is the key to achieve efficient coding in conventional light field coding schemes, where a complicated prediction process is essentially used at the encoder side to exploit the redundancy present in the light field image. We employ Distributed Source Coding (DSC) for light field images, which can extensively lift the computational requirement from the encoding side at the expense of increased computational complexity at the decoder side. The efficiency of DSC is heavily dependent on the quality of side information at the decoder. Therefore, we propose to leverage a learning-based view synthesis method, which takes into account the light field structure to generate high-quality side information. We compare our approach to Distributed Video Coding and Distributed Multi-view Video Coding schemes adapted to the light field framework and relevant standard-based approach, and demonstrate that the proposed view synthesis-based approach can achieve similar performance, while substantially reducing the number of key views to be transmitted.

In this article we present a novel dictionary learning framework designed for compression and sampling of light fields and light field videos. Unlike previous methods, where a single dictionary with one-dimensional atoms is learned, we propose to train a Multidimensional Dictionary Ensemble (MDE). It is shown that learning an ensemble in the native dimensionality of the data promotes sparsity, hence increasing the compression ratio and sampling efficiency. To make maximum use of correlations within the light field data sets, we also introduce a novel nonlocal pre-clustering approach that constructs an Aggregate MDE (AMDE). The pre-clustering not only improves the image quality but also reduces the training time by an order of magnitude in most cases. The decoding algorithm supports efficient local reconstruction of the compressed data, which enables efficient real-time playback of high-resolution light field videos. Moreover, we discuss the application of AMDE for compressed sensing. A theoretical analysis is presented that indicates the required conditions for exact recovery of point-sampled light fields that are sparse under AMDE. The analysis provides guidelines for designing efficient compressive light field cameras. We use various synthetic and natural light field and light field video data sets to demonstrate the utility of our approach in comparison with the state-of-the-art learning-based dictionaries, as well as established analytical dictionaries.

The conventional computer-generated hologram reconstructing photorealistic three-dimensional (3D) images based on ray-wavefront conversion has the disadvantage of spatio-angular resolution trade-off. In this Letter, we propose for the first time, to the best of our knowledge, a computer-generated photorealistic hologram without spatio-angular resolution trade-off based on the additive compressive light field (CLF) approach. The original light field is compressed into multiple layer images through numerical optimization based on the additive light field principle. Then, by independently calculating the wave propagation from each layer image to the hologram plane and adding them together, a CLF hologram is generated. Since the CLF information is presented through a holographic method, the advantage of high resolution in CLF is preserved while the limitation of the number of physically stacked layers (such as liquid crystal displays) is removed, leading to higher quality, larger depth of field, and higher brightness compared with a conventional CLF display. The proposed method is verified with a photorealistic optical experiment.

Light field cameras are considered to have many potential applications since angular and spatial information is captured simultaneously. However, the limited spatial resolution has brought lots of difficulties in developing related applications and becomes the main bottleneck of light field cameras. In this paper, a learning-based method using residual convolutional networks is proposed to reconstruct light fields with higher spatial resolution. The view images in one light field are first grouped into different image stacks with consistent sub-pixel offsets and fed into different network branches to implicitly learn inherent corresponding relations. The residual information in different spatial directions is then calculated from each branch and further integrated to supplement high-frequency details for the view image. Finally, a flexible solution is proposed to super-resolve entire light field images with various angular resolutions. Experimental results on synthetic and real-world datasets demonstrate that the proposed method outperforms other state-of-the-art methods by a large margin in both visual and numerical evaluations. Furthermore, the proposed method shows good performances in preserving the inherent epipolar property in light field images.

Learning an epipolar shift compensation for light field image super-resolution

3D localization of point source is widely used in many fields, such as bioimaging and autonomous driving fields. However, the localization is hard to perform under scattering conditions because of the diffuse effect of the scattering. We propose a novel method for 3D localization of point source under scattering conditions based on light field imaging and deep learning by only one shot. First, we introduce the description of the point source in a light field wise and how to localize a point source by its light field. On the basis, we elaborate on the effect of scattering on a light field and how to retrieve the location of a point source from a light field with scattering. Then, the effect of aberration on a light field will be introduced. We also build an artificial scene and a deep learning framework to perform a 3D localization practically, and the feasibility and accuracy of our method have been evaluated.