KEYWORDS: Performance modeling, Voxels, Data modeling, Quantization, Video compression, Education and training, 3D modeling, Neural networks, Image compression, Video
The Moving Picture Expert Groups (MPEG) recognized that the Neural Radiance Field (NeRF) could serve as a novel image and video compression approach. Therefore, the MPEG established the ad-hoc group called Implicit Neural Visual Representation (INVR) and is currently exploring the potential standardization of 6 Degree of Freedom (6DoF) video compression using NeRF-based technologies. The INVR group is investigating the compression and rendering performance of various NeRF models, including the basic NeRF composed of simple Multi-Layer Perceptrons (MLPs) and the hybrid NeRF that utilizes voxel grids along with MLPs. In this paper, we propose compression methods of basic NeRF and hybrid NeRF using the existing standards, Neural Network Compression (NNC) and Versatile Video Coding (VVC), respectively. The proposed method for the basic NeRF compression utilizes NNC with a network-adaptive bit allocation method. Additionally, the proposed method for compressing the hybrid NeRF, TensoRF, involves transforming the tensor-planes that constitute TensoRF into feature maps and compress them using VVC. In the validation experiments based on the common test conditions (CTCs) defined by INVR, the proposed hybrid NeRF compression demonstrates significantly higher BPP-PSNR performance compared to the state-of-the art method, Vector Quantized Radiance Field (VQRF).
As the need for a video coding technology for a machine that performs intelligent analysis such as object detection, segmentation, and tracking on massive video data has emerged, MPEG is developing a standard called video coding for machines (VCM). VCM is a standard technology for compression of image/video or its features for performing vision tasks of intelligent machines. In this paper, we propose methods that convert multichannel features extracted from an analysis network of input images into a reordered feature map sequence for enhanced compression using VVC. The proposed methods exploit the correlation between channel feature maps using their mean values and sum of absolute difference (SAD) between feature maps in the reordering. Although the proposed methods do not reach the anchor performance of VCM, it shows better coding performance than compressing the feature without channel reordering.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.