State‐of‐the‐Art in Compressed GPU‐Based Direct Volume Rendering

Abstract

Great advancements in commodity graphics hardware have favoured graphics processing unit (GPU)-based volume rendering as the main adopted solution for interactive exploration of rectilinear scalar volumes on commodity platforms. Nevertheless, long data transfer times and GPU memory size limitations are often the main limiting factors, especially for massive, timevarying or multi-volume visualization, as well as for networked visualization on the emerging mobile devices. To address this issue, a variety of level-of-detail (LOD) data representations and compression techniques have been introduced. In order to improve capabilities and performance over the entire storage, distribution and rendering pipeline, the encoding/decoding process is typically highly asymmetric, and systems should ideally compress at data production time and decompress on demand at rendering time. Compression and LOD pre-computation does not have to adhere to real-time constraints and can be performed off-line for high-quality results. In contrast, adaptive real-time rendering from compressed representations requires fast, transient and spatially independent decompression. In this report, we review the existing compressed GPU volume rendering approaches, covering sampling grid layouts, compact representation models, compression techniques, GPU rendering architectures and fast decoding techniques.

Analysis of tensor approximation for compression-domain volume visualization

“…Moreover, as for other compression-domain volume representations, TAs online reconstruction causes only little overhead, exploiting view coherence, GPU acceleration as well as caching. Therefore, TA allows for efficient interactive volume rendering (see also [6]). …”

“…3D data acquisition devices can easily produce data sets too large to be visualized at interactive frame rates at a high quality. Therefore, DVR methods working on compressed 3D data sets -compressiondomain DVR systems -are an ongoing active research topic [6].…”

“…Examples of learned bases are dictionaries which replace the data by a small set of pre-defined and learned codewords, e.g., including vector quantization or sparse coding. For a detailed analysis and references of such compact models, we refer to [6].…”

“…Compression-domain DVR is an ongoing active research area, nevertheless, SVD-like methods have only recently been exploited [6,28,29]. One reason is that the extension of the SVD or principal component analysis (PCA) to higher orders, beyond matrices, is not trivial.…”

“…Moreover, we conducted a performance analysis of the decomposition-reconstruction process. While we are aware that state-of-the-art volume visualization is performed on the GPU [6], we show our performance analysis only on the CPU. In this work, we aimed to illustrate the main conceptual picture on the computation times without going into too many details on parallel implementation versions of the many presented algorithms.…”