Uncompressed Version Research Articles

Fast and Space-Efficient Parallel Algorithms for Influence Maximization

Influence Maximization (IM) is a crucial problem in data science. The goal is to find a fixed-size set of highly influential seed vertices on a network to maximize the influence spread along the edges. While IM is NP-hard on commonly used diffusion models, a greedy algorithm can achieve (1 - 1/ e )-approximation by repeatedly selecting the vertex with the highest marginal gain in influence as the seed. However, we observe two performance issues in the existing work that prevent them from scaling to today's large-scale graphs: space-inefficient memorization to estimate marginal gain, and time-inefficient seed selection process due to a lack of parallelism. This paper significantly improves the scalability of IM using two key techniques. The first is a sketch-compression technique for the independent cascading model on undirected graphs. It allows combining the simulation and sketching approaches to achieve a time-space tradeoff. The second technique includes new data structures for parallel seed selection. Using our new approaches, we implemented PaC-IM : Parallel and Compressed IM. We compare PaC-IM with state-of-the-art parallel IM systems on a 96-core machine with 1.5TB memory. PaC-IM can process the ClueWeb graph with 978M vertices and 75B edges in about 2 hours. On average, across all tested graphs, our uncompressed version is 5--18x faster and about 1.4x more space-efficient than existing parallel IM systems. Using compression further saves 3.8x space with only 70% overhead in time on average.

A digital “flat affect”? Popular speech compression codecs and their effects on emotional prosody

IntroductionCalls via video apps, mobile phones and similar digital channels are a rapidly growing form of speech communication. Such calls are not only— and perhaps less and less— about exchanging content, but about creating, maintaining, and expanding social and business networks. In the phonetic code of speech, these social and emotional signals are considerably shaped by (or encoded in) prosody. However, according to previous studies, it is precisely this prosody that is significantly distorted by modern compression codecs. As a result, the identification of emotions becomes blurred and can even be lost to the extent that opposing emotions like joy and anger or disgust and sadness are no longer differentiated on the recipients' side. The present study searches for the acoustic origins of these perceptual findings.MethodA set of 108 sentences from the Berlin Database of Emotional Speech served as speech material in our study. The sentences were realized by professional actors (2m, 2f) with seven different emotions (neutral, fear, disgust, joy, boredom, anger, sadness) and acoustically analyzed in the original uncompressed (WAV) version and as well as in strongly compressed versions based on the four popular codecs AMR-WB, MP3, OPUS, and SPEEX. The analysis included 6 tonal (i.e. f0-related) and 7 non-tonal prosodic parameters (e.g., formants as well as acoustic-energy and spectral-slope estimates).ResultsResults show significant, codec-specific distortion effects on all 13 prosodic parameter measurements compared to the WAV reference condition. Means values of automatic measurement can, across sentences, deviate by up to 20% from the values of the WAV reference condition. Moreover, the effects go in opposite directions for tonal and non-tonal parameters. While tonal parameters are distorted by speech compression such that the acoustic differences between emotions are increased, compressing non-tonal parameters make the acoustic-prosodic profiles of emotions more similar to each other, particularly under MP3 and SPEEX compression.DiscussionThe term “flat affect” comes from the medical field and describes a person's inability to express or display emotions. So, does strong compression of emotional speech create a “digital flat affect”? The answer to this question is a conditional “yes”. We provided clear evidence for a “digital flat affect”. However, it seems less strongly pronounced in the present acoustic measurements than in previous perception data, and it manifests itself more strongly in non-tonal than in tonal parameters. We discuss the practical implications of our findings for the everyday use of digital communication devices and critically reflect on the generalizability of our findings, also with respect to their origins in the codecs' inner mechanics.

Fuzzy Logic-Based Adaptive Point Cloud Video Streaming

Point cloud video provides 6 degrees of freedom (6DoF) viewing experiences to allow users to freely select the viewing angles of 3D scenes and is expected to be the next-generation video. This paper studies the point cloud video streaming and proposes a fuzzy logic-based point cloud video streaming scheme to solve the inherent technical issues. In particular, a point cloud video is first partitioned into smaller tiles, along with a low-quality base layer covering the entire video. Each tile is encoded into different quality levels, and both the compressed and uncompressed (i.e., decoded) versions of each tile are prepared for selection. Then, based on the user's viewing angle and predicted future network bandwidth condition, fuzzy logic empowered quality level selection, with properly defined novel fuzzification, fuzzy rules, and defuzzification, is conducted to maximize the received point cloud video quality under the communication resource, computational resource and quality requirements constraints. Extensive simulations based on real point cloud video sequences and network traces are conducted, and the results reveal the superiority of the proposed scheme over the baseline scheme. To the best of our knowledge, this is the first work studying point cloud video streaming using fuzzy logic.

A multi-level descriptor using ultra-deep feature for image retrieval

CNN(Convolution Neural Network)-based descriptor generation is extensively studied recently for image retrieval. CNN deep feature trained for image classification is proved to have good transferability for image retrieval task. However, building a highly discriminative descriptor with CNN feature is still an important issue. The feature of the fully-connected layer is usually used and the shallow features of an image are ignored. In this paper, we proposed a simple and effective multi-level descriptor. Firstly, we proposed a multi-level feature fusion (MFF) method to capture low-level color/texture and high-level semantic information simultaneously. MFF replaces the commonly-used “object-level” with “part-level”, and the filters of convolution layer are seen as part detectors, instead of using an object detector method explicitly. The complementary nature of low-level and high-level feature benefits MFF greatly. Secondly, we trained a neural net with class information to further improve the discriminative power of MFF. Our MFF achieves good performance on public image retrieval datasets. Finally, a compressed version is proposed and achieves close performance to the uncompressed version.

Influence of acquisition frame-rate and video compression techniques on pulse-rate variability estimation from vPPG signal

In this paper, common time- and frequency-domain variability indexes obtained by pulse rate variability (PRV) series extracted from video-photoplethysmographic signal (vPPG) were compared with heart rate variability (HRV) parameters calculated from synchronized ECG signals. The dual focus of this study was to analyze the effect of different video acquisition frame-rates starting from 60 frames-per-second (fps) down to 7.5 fps and different video compression techniques using both lossless and lossy codecs on PRV parameters estimation. Video recordings were acquired through an off-the-shelf GigE Sony XCG-C30C camera on 60 young, healthy subjects (age 23±4 years) in the supine position. A fully automated, signal extraction method based on the Kanade-Lucas-Tomasi (KLT) algorithm for regions of interest (ROI) detection and tracking, in combination with a zero-phase principal component analysis (ZCA) signal separation technique was employed to convert the video frames sequence to a pulsatile signal. The frame-rate degradation was simulated on video recordings by directly sub-sampling the ROI tracking and signal extraction modules, to correctly mimic videos recorded at a lower speed. The compression of the videos was configured to avoid any frame rejection caused by codec quality leveling, FFV1 codec was used for lossless compression and H.264 with variable quality parameter as lossy codec. The results showed that a reduced frame-rate leads to inaccurate tracking of ROIs, increased time-jitter in the signals dynamics and local peak displacements, which degrades the performances in all the PRV parameters. The root mean square of successive differences (RMSSD) and the proportion of successive differences greater than 50 ms (PNN50) indexes in time-domain and the low frequency (LF) and high frequency (HF) power in frequency domain were the parameters which highly degraded with frame-rate reduction. Such a degradation can be partially mitigated by up-sampling the measured signal at a higher frequency (namely 60 Hz). Concerning the video compression, the results showed that compression techniques are suitable for the storage of vPPG recordings, although lossless or intra-frame compression are to be preferred over inter-frame compression methods. FFV1 performances are very close to the uncompressed (UNC) version with less than 45% disk size. H.264 showed a degradation of the PRV estimation directly correlated with the increase of the compression ratio.

Ratat1: A Digital Rat Brain Stereotaxic Atlas Derived from High-Resolution MRI Images Scanned in Three Dimensions.

A digital rat brain atlas, in PDF format, was derived from high-resolution in situ magnetic resonance images (MRI). The images depict transverse, sagittal, and horizontal views in contiguous 0.2 mm slices, with a planar resolution of 50 µm. Each slice was indexed to the skull surface as well as the skull landmarks Bregma and Lambda. Using a common interactive file format enables users to customize the images for their application, including image reproduction with or without background or labels. The objective was to provide an open access supplementary guide for stereotaxic neurosurgery that permits more accurate visualization of targets and determination of target coordinates in Long Evans rats. A secondary objective was to present the atlas in the widely used PDF format, not requiring specialized software for viewing, copying or unpacking. A compressed version of the atlas is available as a supplement, while an uncompressed version is available at: https://figshare.com/articles/RatAtlas_2_0_Locked_pdf/3144955.

Objective and subjective evaluation of High Dynamic Range video compression

A number of High Dynamic Range (HDR) video compression algorithms proposed to date have either been developed in isolation or only-partially compared with each other. Previous evaluations were conducted using quality assessment error metrics, which for the most part were developed for qualitative assessment of Low Dynamic Range (LDR) videos. This paper presents a comprehensive objective and subjective evaluation conducted with six published HDR video compression algorithms. The objective evaluation was undertaken on a large set of 39 HDR video sequences using seven numerical error metrics namely: PSNR, logPSNR, puPSNR, puSSIM, Weber MSE, HDR-VDP and HDR-VQM. The subjective evaluation involved six short-listed sequences and two ranking-based subjective experiments with hidden reference at two different output bitrates with 32 participants each, who were tasked to rank distorted HDR video footage compared to an uncompressed version of the same footage. Results suggest a strong correlation between the objective and subjective evaluation. Also, non-backward compatible compression algorithms appear to perform better at lower output bit rates than backward compatible algorithms across the settings used in this evaluation.

Longest Common Subsequence Problem for Run-Length-Encoded Strings

In this paper, we present a new and efficient algorithm for solving the Longest Common Subsequence (LCS) problem between two run-length-encoded (RLE) strings. Suppose Ŷ and ^X are two RLE strings having length ^k and ^l respectively. Also assume that Y and X are the two uncompressed versions of the two RLE strings Ŷ and ^X having length k and ` respectively. Then, our algorithm runs in O((^k+ ^l)+Rlog log(^k^l)+Rloglogω) time, where ω = k + l and R is the total number of ordered pairs of positions at which the two RLE strings match. Our algorithm outperforms the best algorithms for the same problem in the literature.

Acoustical correlates of performance on a dynamic range compression discrimination task

Dynamic range compression is widely used to reduce the difference between the most and least intense portions of a signal. Such compression distorts the shape of the amplitude envelope of a signal, but it is unclear to what extent such distortions are actually perceivable by listeners. Here, the ability to distinguish between compressed and uncompressed versions of a noise vocoded sentence was initially measured in listeners with normal hearing while varying the threshold, ratio, attack, and release parameters. This narrow condition was selected in order to characterize perception under the most favorable listening conditions. The average behavioral sensitivity to compression was highly correlated to several acoustical indices of modulation depth. In particular, performance was highly correlated to the Euclidean distance between the modulation spectra of the uncompressed and compressed signals. Suggesting that this relationship is not restricted to the initial test conditions, the correlation remained largely unchanged both (1) when listeners with normal hearing were tested using a time-compressed version of the original signal, and (2) when listeners with impaired hearing were tested using the original signal. If this relationship generalizes to more ecologically valid conditions, it will provide a straightforward method for predicting the detectability of compression-induced distortions.

JPEG vs. JPEG2000: benchmarking with dermatological images

Despite the importance of images in the discipline and the diffusion of digital imaging devices, the issue of image compression in dermatology was discussed only in few studies, which yielded results often not comparable, and left some unanswered questions. To evaluate and compare the performance of the JPEG and JPEG2000 algorithms for compression of dermatological images. Nineteen macroscopic and fifteen videomicroscopic images of skin lesions were compressed with JPEG and JPEG2000 at 18 different compression rates, from 90% to 99.5%. Compressed images were shown, next to uncompressed versions, to three dermatologists with different experience, who judged quality and suitability for educational/scientific and diagnostic purposes. Moreover, alterations and quality were evaluated by calculation of mean 'distance' of pixel colors between compressed and original images and by peak signal-to-noise ratio, respectively. JPEG2000 was qualitatively better than JPEG at all compression rates, particularly highest ones, as shown by dermatologists' ratings and objective parameters. Agreement between raters was high, but with some differences in specific cases, showing that different professional experience can influence judgement on images. In consideration of its high qualitative performance and wide diffusion, JPEG2000 represents an optimal solution for the compression of digital dermatological images.

Compressing pathology whole-slide images using a human and model observer evaluation

Introduction: We aim to determine to what degree whole-slide images (WSI) can be compressed without impacting the ability of the pathologist to distinguish benign from malignant tissues. An underlying goal is to demonstrate the utility of a visual discrimination model (VDM) for predicting observer performance. Materials and Methods: A total of 100 regions of interest (ROIs) from a breast biopsy whole-slide images at five levels of JPEG 2000 compression (8:1, 16:1, 32:1, 64:1, and 128:1) plus the uncompressed version were shown to six pathologists to determine benign versus malignant status. Results: There was a significant decrease in performance as a function of compression ratio (F = 14.58, P < 0.0001). The visibility of compression artifacts in the test images was predicted using a VDM. Just-noticeable difference (JND) metrics were computed for each image, including the mean, median, ≥90th percentiles, and maximum values. For comparison, PSNR (peak signal-to-noise ratio) and Structural Similarity (SSIM) were also computed. Image distortion metrics were computed as a function of compression ratio and averaged across test images. All of the JND metrics were found to be highly correlated and differed primarily in magnitude. Both PSNR and SSIM decreased with bit rate, correctly reflecting a loss of image fidelity with increasing compression. Observer performance as measured by the Receiver Operating Characteristic area under the curve (ROC Az) was nearly constant up to a compression ratio of 32:1, then decreased significantly for 64:1 and 128:1 compression levels. The initial decline in Az occurred around a mean JND of 3, Minkowski JND of 4, and 99th percentile JND of 6.5. Conclusion: Whole-slide images may be compressible to relatively high levels before impacting WSI interpretation performance. The VDM metrics correlated well with artifact conspicuity and human performance.

Fast and flexible word searching on compressed text

We present a fast compression technique for natural language texts. The novelties are that (1) decompression of arbitrary portions of the text can be done very efficiently, (2) exact search for words and phrases can be done on the compressed text directly, using any known sequential pattern-matching algorithm, and (3) word-based approximate and extended search can also be done efficiently without any decoding. The compression scheme uses a semistatic word-based model and a Huffman code where the coding alphabet is byte-oriented rather than bit-oriented. We compress typical English texts to about 30% of their original size, against 40% and 35% forCompressandGzip, respectively. Compression time is close to that ofCompressand approximately half of the time ofGzip, and decompression time is lower than that ofGzipand one third of that ofCompress. We present three algorithms to search the compressed text. They allow a large number of variations over the basic word and phrase search capability, such as sets of characters, arbitrary regular expressions, and approximate matching. Separators and stopwords can be discarded at search time without significantly increasing the cost. When searching for simple words, the experiments show that running our algorithms on a compressed text is twice as fast as running the best existing software on the uncompressed version of the same text. When searching complex or approximate patterns, our algorithms are up to 8 times faster than the search on uncompressed text. We also discuss the impact of our technique in inverted files pointing to logical blocks and argue for the possibility of keeping the text compressed all the time, decompressing only for displaying purposes.

MLP for adaptive postprocessing block-coded images

A new technique based on the multilayer perceptron (MLP) neural network is proposed for blocking-artifact removal in block-coded images. The new method is based on the concept of learning-by-examples. The compressed image and its original uncompressed version are used to train the neural networks. In the developed scheme, inter-block slopes of the compressed image are used as input, the difference between the original uncompressed and the compressed image is used as desired output for training the networks. Blocking-artifact removal is realized by adding the neural network's outputs to the compressed image. The new technique has been applied to process JPEG compressed images. Experimental results show significant improvements in both visual quality and peak signal-to-noise ratio. It is also shown the present method is comparable to other state of the art techniques for quality enhancement in block-coded images.

A pyramid-based error-bounded encoder: An evaluation on X-ray chest images

A novel encoder based on an enhanced Laplacian pyramid is proposed for compression of Medical grey-scale images: major details are prioritized through an adaptive decision rule embedded in a uniform threshold quantizer with noise feedback. The major benefit of this content-driven feedback quantizer is that significant features are straightforwardly propagated throughout the pyramid, thus enhancing compactness and visual quality of the reduced-resolution versions progressively associated with the code stream. Nevertheless, the reconstruction error is determined only by the size of the quantization step at the base level of the pyramid, thereby making it possible for the maximum absolute reconstruction error to be easily and strongly upperbounded ( near-lossless compression), as often required in archiving medical images, for diagnostic and legal purposes. Both lossless and lossy coding show favorable comparisons with JPEG in objective terms, i.e., compression ratios and distortion plots. Lossy coding outperforms JPEG also subjectively, due to the absence of visual impairments and diagnostic artifacts even at very low rates. This feature is also evidenced in a preliminary ROC analysis on a set of X-ray chest images. The pyramid encoder produces compressed images whose diagnostic quality seems to be comparable, for medium rates, to that of the uncompressed versions, and superior to that of the JPEG coded versions.