Lossy Compression Performance Research Articles

Black-box statistical prediction of lossy compression ratios for scientific data

Lossy compressors are increasingly adopted in scientific research, tackling volumes of data from experiments or parallel numerical simulations and facilitating data storage and movement. In contrast with the notion of entropy in lossless compression, no theoretical or data-based quantification of lossy compressibility exists for scientific data. Users rely on trial and error to assess lossy compression performance. As a strong data-driven effort toward quantifying lossy compressibility of scientific datasets, we provide a statistical framework to predict compression ratios of lossy compressors. Our method is a two-step framework where (i) compressor-agnostic predictors are computed and (ii) statistical prediction models relying on these predictors are trained on observed compression ratios. Proposed predictors exploit spatial correlations and notions of entropy and lossyness via the quantized entropy. We study 8+ compressors on 6 scientific datasets and achieve a median percentage prediction error less than 12%, which is substantially smaller than that of other methods while achieving at least a 8.8× speedup for searching for a specific compression ratio and 7.8× speedup for determining the best compressor out of a collection.

Spatiotemporal Prediction Based on Feature Classification for Multivariate Floating-Point Time Series Lossy Compression

A large amount of time series is produced because of the frequent use of IoT devices and sensors. Time series compression is widely adopted to reduce storage overhead and transport costs. At present, most state-of-the-art approaches focus on univariate time series. Therefore, the task of compressing multivariate time series (MTS) is still an important but challenging problem. Traditional MTS compression methods treat each variable individually, ignoring the correlations across variables. This paper proposes a novel MTS prediction method, which can be applied to compress MTS to achieve a higher compression ratio. The method can extract the spatial and temporal correlation across multiple variables, achieving a more accurate prediction and improving the lossy compression performance of MTS based on the prediction-quantization-entropy framework. We use a convolutional neural network (CNN) to extract the temporal features of all variables within the window length. Then the features generated by CNN are transformed, and the image classification algorithm extracts the spatial features of the transformed data. Predictions are made according to spatiotemporal characteristics. To enhance the robustness of our model, we integrate the AR autoregressive linear model in parallel with the proposed network. Experimental results demonstrate that our work can improve the prediction accuracy of MTS and the MTS compression performance in most cases.

A new, enhanced EZW image codec with subband classification

In this paper, an enhanced version of Embedded zerotree wavelet (EZW) image coding algorithm is proposed, referred to as EZW-SC. By exploiting a new principle that relies on a subband classification concept, the enhanced algorithm allows the prediction of insignificant subbands at early passes, along with the use of an improved significance map. This reduces the redundancy of zerotree symbols, speeds up the coding process and improves the coding of significant coefficients. In fact, the EZW-SC algorithm scans only significant subbands and significantly improves the lossy compression performance with the conventional EZW. Moreover, new EZW-based schemes are presented to perform colour image coding by taking advantage of the interdependency of the colour components. Experimental results show clear superiority of the proposed algorithms over the conventional EZW as well as other related EZW schemes at various bit rates in both greyscale and colour image compression.

Lossless to Lossy Dual-Tree BEZW Compression for Hyperspectral Images

This paper proposes a lossless to lossy compression scheme for hyperspectral images based on dual-tree Binary Embedded Zerotree Wavelet (BEZW) algorithm. The algorithm adapts Karhunen-Loève Transform and Discrete Wavelet Transform to achieve 3-D integer reversible hybrid transform and decorrelate spectral and spatial data. Since statistics of the hyperspectral image are not symmetrical, the asymmetrical dual-tree structure is introduced. The 3-D BEZW algorithm compresses hyperspectral images by implementing progressive bitplane coding. The lossless and lossy compression performance is compared with other state-of-the-art predictive coding and transform-based coding algorithms on Airborne Visible/Infrared Imaging Spectrometer images. Results show that the 3-D-BEZW lossless compression performance is comparable with the best predictive algorithms, while its computational cost is comparable with those of transform-based algorithms.

JPEG2000 and JPEG: Image Quality Comparison of Compressed Medical Modalities

Due to the constrained bandwidth and storage capacity, medical images must be compressed before transmission and storage. However, the compression will reduce the image fidelity, especially when the image is compressed of lower bit rate, which cannot be tolerated in medical field. In this paper, the lossy compression performance of the new JPEG2000 and the more conventional Joint Photographic Experts Group (JPEG ) is studied and compared. The parameters used for comparison include the compression efficiency, peak signal-to-noise ratio (PSNR), picture quality scale (PQS), and mean opinion score (MOS). Four types of medical modalities are used: magnetic resonance imaging (MRI), ultrasound, X-ray and computed tomography (CT). Overall, the study shows that JPEG2000 compression is more acceptable and superior compared to JPEG. The compression threshold point is also roughly determined for each of the medical image type.

Three-dimensional medical image compression based on low-complexity RKLT

A low-complexity reversible integer Karhunen-Loeve transform (low-RKLT) is proposed for three-dimensional (3D) medical image compression. Low-RKLT is used for exploiting the correlation existing among neighbouring slices followed by integer wavelet transform in the spatial domain. Simulation results show that low-RKLT provides comparable lossless compression performance and obviously better lossy compression performance compared with integer wavelet transform.

Hyperspectral image compression using a three-dimensional embedded zeroblock coding algorithm

Motion-compensated three-dimensional embedded zeroblock coding (MC 3-D EZBC) is a successful state-of-the-art video compression algorithm. We propose a hyperspectral image compression coder based on the 3-D EZBC algorithm without motion compensation. This coder adopts the 3-D wavelet transform to decorrelate and the 3-D EZBC algorithm without motion compensation to process bitplane zeroblock coding. For achieving good coding performance, the diverse 3-D wavelet transform structures and the several wavelet filters are respectively compared and evaluated on the basis of floating-point lossy compression and lossless-to-lossy compression. We also study the problems of the optimal unitary scaling factors and list initialization order. Finally, the best choices were found for a given application, via the extensive experiments and analyses. Moreover, in comparison with several state-of-the-art wavelet coding algorithms, 3-D EZBC can provide better compression performance and unsupervised classification accuracy. Experimental results show that the average lossy compression performance (in floating-point mode and integer-based mode) of our coder respectively outperforms 3-D set partitioning in hierarchical trees (SPIHT) by 1.26 dB, 3-D set-partitioned embedded block (SPECK) by 0.68 dB, symmetric-tree (AT) 3-D SPIHT by 0.39 dB, and JPEG 2000-MC by 0.25 dB at 0.1 to 3.0 bits per pixel per band, and the lossless coding performance of 3-D EZBC is about 5% to 7% better than that of 3-D SPECK, 3-D SPIHT, and AT 3-D SPIHT. So the 3-D EZBC algorithm is also a good candidate to compress hyperspectral images.

Motion Compensated Lossy-to-Lossless Compression of 4-D Medical Images Using Integer Wavelet Transforms

This paper proposes a method for progressive lossy-to-lossless compression of four-dimensional (4-D) medical images (sequences of volumetric images over time) by using a combination of three-dimensional (3-D) integer wavelet transform (IWT) and 3-D motion compensation. A 3-D extension of the set-partitioning in hierarchical trees (SPIHT) algorithm is employed for coding the wavelet coefficients. To effectively exploit the redundancy between consecutive 3-D images, the concepts of key and residual frames from video coding is used. A fast 3-D cube matching algorithm is employed to do motion estimation. The key and the residual volumes are then coded using 3-D IWT and the modified 3-D SPIHT. The experimental results presented in this paper show that our proposed compression scheme achieves better lossy and lossless compression performance on 4-D medical images when compared with JPEG-2000 and volumetric compression based on 3-D SPIHT.

LOSSILY COMPRESSED MEDICAL MODALITIES: OBJECTIVE AND SUBJECTIVE QUALITY MEASURES IN RELEVANT TO THE STATISTICAL AND FREQUENCY EVALUATIONS

Compressed image will always reduce the image fidelity. This is especially so if the image is compressed at lower bit rates, which most of the time cannot be tolerated. In medical field, this is even more significant. But compression is necessary due to the limited transmission bandwidth and storage capacity. In this paper, the lossy compression performance of the new JPEG2000 and the more conventional JPEG is being studied. The common objective measure, peak signal-to-noise ratio (PSNR) is definitely not adequate as a perceptual quality measurement. Therefore, subjective measures such as picture quality scale (PQS) and mean opinion score (MOS) are included, plus the statistical and frequency measures for the image characteristics evaluation. Four types of medical modalities are used — X-Ray, Magnetic Resonance Imaging (MRI), Ultrasound, and Computed Tomography (CT). From these various measurements, the general acceptance compression bit rate for each of the medical modality is determined.

Optimization and implementation of the integer wavelet transform for image coding

This paper deals with the design and implementation of an image transform coding algorithm based on the integer wavelet transform (IWT). First of all, criteria are proposed for the selection of optimal factorizations of the wavelet filter polyphase matrix to be employed within the lifting scheme. The obtained results lead to the IWT implementations with very satisfactory lossless and lossy compression performance. Then, the effects of finite precision representation of the lifting coefficients on the compression performance are analyzed, showing that, in most cases, a very small number of bits can be employed for the mantissa keeping the performance degradation very limited. Stemming from these results, a VLSI architecture is proposed for the IWT implementation, capable of achieving very high frame rates with moderate gate complexity.

Quantisation step selection schemes in JPEG2000

Modified quantisation step selection schemes for the uniform dead-zone scalar quantisation as used in Part I of the JPEG2000 standard are presented. Effects of these quantisation schemes on the lossy compression performance of 5/3 integer wavelet transforms are evaluated. The experiment is carried out in the JPEG2000 framework. Similar consideration may be applied to the real wavelet transforms for lossy image compression.

Reversible integer-to-integer wavelet transforms for image compression: performance evaluation and analysis

In the context of image coding, a number of reversible integer-to-integer wavelet transforms are compared on the basis of their lossy compression performance, lossless compression performance, and computational complexity. Of the transforms considered, several were found to perform particularly well, with the best choice for a given application depending on the relative importance of the preceding criteria. Reversible integer-to-integer versions of numerous transforms are also compared to their conventional (i.e., nonreversible real-to-real) counterparts for lossy compression. At low bit rates, reversible integer-to-integer and conventional versions of transforms were found to often yield results of comparable quality. Factors affecting the compression performance of reversible integer-to-integer wavelet transforms are also presented, supported by both experimental data and theoretical arguments.