Search

Instead of being observed by human, multimedia data are now more and more fed into machines to perform different kinds of semantic analysis. One image may be analyzed multiple times by different machine vision algorithms for different purposes. While machine vision-oriented image compression has been studied, the existing methods are usually driven by a specific machine vision task, and may not be applicable for other tasks. We address the task-generic image compression, in the hope that an image is compressed once but used multiple times for different tasks, all with satisfactory performance. Our study is based on the end-to-end learned image compression. We focus ourselves on the distortion metric, i.e., finding out a task-agnostic metric to estimate the quality of reconstructed images. On the one hand, we study deep feature distance as the metric, which transforms images into a latent space by a pretrained convolutional network -- the latent space is believed to be more aligned to semantics -- and calculates distance in the latent space. On the other hand, inspired by the saliency mechanism, we study an importance-weighted pixel distance as the metric, where the weights are generated to reflect the importance of the pixels to semantics. Moreover, we combine the two distances into one metric to investigate their complementary nature. An extensive set of experiments are performed to evaluate these metrics. Experimental results show that the combined metric performs the best, and leads to 20.79%~42.69% bits saving under the same semantic analysis performance, compared to the same network but optimized for signal fidelity. Interestingly, we observe that using the combined metric also improves the visual quality of the reconstructed images.

Computer vision is nowadays is one of the promising & evolving areas in information technology. Image processing is increasingly used in several applications such as automotive, medical or aerospace. Every computer vision application involves capturing several images with the help of camera. Image acquisition, image transmission & image security are the three important aspects of image processing system. This paper includes development of complete image processing system including improved median filtering with better PSNR & operating frequency, Image compression module with better PSNR & compression ratio, Image Encryption module using Advanced Encryption Standard.

Multi-camera networks are increasingly becoming pervasive in many monitoring and surveillance applications, and have attracted much attention in distributed systems with collaborative, real-time decision-making capabilities. While in-network data compression brings significant energy savings in camera nodes, signal representation using sparse approximations and overcomplete dictionaries have been shown to outperform traditional compression methods. In this work, an end-to-end and real-time solution is designed and implemented to enable energy-efficient and robust dictionary learning in distributed camera networks by leveraging the spatial correlation of the collected multimedia data. Traditional distributed dictionary learning relies on consensus-building algorithms, which involve communicating with neighboring nodes until convergence is achieved. Existing methods, however, do not exploit spatial correlations in camera networks for improved energy efficiency. In contrast, low-computational-complexity metrics are employed in this work to quantify and exploit the spatial correlation across camera nodes in a wireless network for efficient distributed dictionary learning and in-network image compression. The performance of the proposed approach is validated through extensive simulations on public datasets as well as via real-world experiments on a testbed composed of Raspberry Pi nodes.

Over the years, photometric three-dimensional (3D) reconstruction gained increasing importance in several disciplines, especially in cultural heritage preservation. While increasing sizes of images and datasets enhanced the overall reconstruction results, requirements in storage got immense. Additionally, unsharp areas in the background have a negative influence on 3D reconstructions algorithms. Handling the sharp foreground differently from the background simultaneously helps to reduce storage size requirements and improves 3D reconstruction results. In this article, we examine regions outside the Depth of Field (DoF) and eliminate their inaccurate information to 3D reconstructions. We extract DoF maps from the images and use them to handle the foreground and background with different compression backends, making sure that the actual object is compressed losslessly. Our algorithm achieves compression rates between 1:8 and 1:30 depending on the artifact and DoF size and improves the 3D reconstruction.

Transferring images in a wireless multimedia sensor network (WMSN) knows a fast development in both research and fields of application. Nevertheless, this area of research faces many problems such as the low quality of the received images after their decompression, the limited number of reconstructed images at the base station, and the high-energy consumption used in the process of compression and decompression. In order to fix these problems, we proposed a compression method based on the classic discrete wavelet transform (DWT). Our method applies the wavelet compression technique multiple times on the same image. As a result, we found that the number of received images is higher than using the classic DWT. In addition, the quality of the received images is much higher compared to the standard DWT. Finally, the energy consumption is lower when we use our technique. Therefore, we can say that our proposed compression technique is more adapted to the WMSN environment.

With the advancements in modern medical imaging systems, the diagnostic image data has increased exponentially. The future medical applications seek medical imaging devices to be portable. Image quality and real-time processing are of prime importance in medical image compression. Therefore the volumetric medical images need to be compressed in perceptually lossless manner. Also the time taken to compress the images before transmitting (or storing), should be small. In this paper, an algorithm for lossless and perceptually-lossless medical image compression is proposed. The proposed algorithm uses two small lists and two small state-tables to encode an image. The compression efficiency is comparable to the state-of-the-art lossless compression techniques. Also the computational complexity and memory requirement are realistic for portable medical imaging devices. Combining all the three features, it is obvious that the proposed algorithm is a better candidate for image compression in comparison to all the state-of-art compression algorithms that we know of, for volumetric-medical imaging systems.

Recent works on learned image compression perform encoding and decoding processes in a full-resolution manner, resulting in two problems when deployed for practical applications. First, parallel acceleration of the autoregressive entropy model cannot be achieved due to serial decoding. Second, full-resolution inference often causes the out-of-memory(OOM) problem with limited GPU resources, especially for high-resolution images. Block partition is a good design choice to handle the above issues, but it brings about new challenges in reducing the redundancy between blocks and eliminating block effects. To tackle the above challenges, this paper provides a learned block-based hybrid image compression (LBHIC) framework. Specifically, we introduce explicit intra prediction into a learned image compression framework to utilize the relation among adjacent blocks. Superior to context modeling by linear weighting of neighbor pixels in traditional codecs, we propose a contextual prediction module (CPM) to better capture long-range correlations by utilizing the strip pooling to extract the most relevant information in neighboring latent space, thus achieving effective information prediction. Moreover, to alleviate blocking artifacts, we further propose a boundary-aware postprocessing module (BPM) with the edge importance taken into account. Extensive experiments demonstrate that the proposed LBHIC codec outperforms the VVC, with a bit-rate conservation of 4.1%, and reduces the decoding time by approximately 86.7% compared with that of state-of-the-art learned image compression methods., 13 pages, 13 figures, accepted by IEEE Trans. on Circuits and Systems for Video Technology

With the emergence of light field imaging in recent years, the compression of its elementary image array (EIA) has become a significant problem. Our coding framework includes modeling and reconstruction. For the modeling, the covariance-matrix form of the 4-D Epanechnikov kernel (4-D EK) and its correlated statistics were deduced to obtain the 4-D Epanechnikov mixture models (4-D EMMs). A 4-D Epanechnikov mixture regression (4-D EMR) was proposed based on this 4-D EK, and a 4-D adaptive model selection (4-D AMLS) algorithm was designed to realize the optimal modeling for a pseudo video sequence (PVS) of the extracted key-EIA. A linear function based reconstruction (LFBR) was proposed based on the correlation between adjacent elementary images (EIs). The decoded images realized a clear outline reconstruction and superior coding efficiency compared to high-efficiency video coding (HEVC) and JPEG 2000 below approximately 0.05 bpp. This work realized an unprecedented theoretical application by (1) proposing the 4-D Epanechnikov kernel theory, (2) exploiting the 4-D Epanechnikov mixture regression and its application in the modeling of the pseudo video sequence of light field images, (3) using 4-D adaptive model selection for the optimal number of models, and (4) employing a linear function-based reconstruction according to the content similarity.

A 3D mesh can be reconstructed from multiple viewpoint images or from a single structured light image. Lossy compression of such images by standard techniques such as JPEG at high compression ratios lead to 3D reconstruction being adversely affected by artifacts and missing vertices. In this paper we demonstrate an improved algorithm capable of high compression ratios without adversely affecting 3D reconstruction and with minimum data loss. The compression algorithm starts by applying block DCT over the input image, and the transformed data being quantized using an optimized quantization matrix. The quantized coefficients of each block are arranged as a 1D array and saved with other block’s data in a larger matrix of coefficients. The DC coefficients are subject to a first order difference whose values are referred to as residual array. The AC coefficients are reduced by eliminating zeros and saving the non-zero values in a reduced coefficients array using a mask of 0 (for a block of zeros) and 1 (for a block of non-zeros). Finally, arithmetic coding is applied to both coefficients and residual arrays. At decompression stage, the coefficients matrix is regenerated by scanning the coefficients array and examining the headers to substitute zero and non-zero data. This matrix is then added to the residual array to obtain the original DC values. The IDCT is then applied to obtain the original image. The proposed algorithm has been tested with images of varying sizes in the context of 3D reconstruction. Results demonstrate that our proposed algorithm is superior to traditional JPEG at higher compression ratios with high perceptual quality of images and the ability to reconstruct the 3D models more effectively, both for structured light images and for sequences of multiple viewpoint images.

In this paper, an efficient technique for image compression and quality retrieval using matrix completion is presented. The proposed technique is based on low-rank matrix completion using singular value truncation and thresholding. Here, an image is decomposed using singular value decomposition (SVD) to obtain a low rank of image data, which is approximated in compressed form. Later on, singular value thresholding algorithm is exploited to retrieve visual quality of the compressed image. The presented method is easily applicable for various visual characteristics of the image for different compression efficiency. A detailed analysis has been presented to show the efficiency of proposed method in term of compression as well as quality retrieval. It is evident from experimental results that a maximum of 80% compression is achieved with acceptable visual quality as per human vision system (HVS).

In mathematical theory, the discrete cosine transform (DCT) is a lossless orthogonal transformation method which means it outputs exactly the same values of the input after the inverse transformation. However, this is impossible in today’s technology due to the limited capacity of processors in which the maximum value that a number can take is 2 64- 1 (20-digit number) in a 64-bit register. Since the DCT employs the floating values higher than this precision, there occurs a round-off error which causes a particular loss of information after the inverse transformation. For this reason, the dynamic range of the DCT coefficients should be reduced so that fewer precision digits are employed in the DCT calculations, thereby the round-off error and loss of information are minimized. In this study, conventional DCT equations are improved both in forward and inverse transformation for the sake of high-performance JPEG image compression. The proposed method reduces the dynamic range of the DCT coefficients and provides a low dynamic range DCT (LDR-DCT) by weighting the DCT coefficients with respect to the frequency level. The effectiveness of the proposed LDR-DCT method is experimented mainly by observing the inter-correlation between the compression ratio and the peak signal-to-noise ratio (PSNR) values which is defined as the compression performance (CP). An extensive experimental benchmarking study is done using the publicly available KODAK image dataset in both grayscale and RGB color spaces, separately. According to the experimental results, the average compression performance (CP) is increased up to about 26% in grayscale images and about 17% in RGB images when the quantization factors (21–121) are employed in the quantization process. Additionally, it is observed that there is an average increment in the compression performance (CP) up to about 8% in grayscale images and about 7% in RGB images when the standard IrfanView quantization tables (quality level of 40 to the quality level of 90) are applied. On the other hand, in the absence of quantization when either the quantization factor of 1 or the standard IrfanView quantization table with the quality level of 100 is applied, it is also observed that there is an average increment in the PSNR value up to about 15% in grayscale images and about 33% in RGB images with respect to the average PSNR values of 24 images in the KODAK image dataset. Therefore, though the proposed LDR-DCT method without quantization does not change the compression ratio, it improves the quality of the output obtained after the inverse transform dramatically. In other words, the conventional DCT method should be replaced by the proposed LDR-DCT method in certain areas where compression is not required. Besides, the study claims that the proposed LDR-DCT method can provide at least the same JPEG image quality as the conventional DCT method with much higher compression ratios if the quantization tables are redesigned accordingly. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

Built on deep networks, end-to-end optimized image compression has made impressive progress in the past few years. Previous studies usually adopt a compressive auto-encoder, where the encoder part first converts image into latent features, and then quantizes the features before encoding them into bits. Both the conversion and the quantization incur information loss, resulting in a difficulty to optimally achieve arbitrary compression ratio. We propose iWave++ as a new end-to-end optimized image compression scheme, in which iWave, a trained wavelet-like transform, converts images into coefficients without any information loss. Then the coefficients are optionally quantized and encoded into bits. Different from the previous schemes, iWave++ is versatile: a single model supports both lossless and lossy compression, and also achieves arbitrary compression ratio by simply adjusting the quantization scale. iWave++ also features a carefully designed entropy coding engine to encode the coefficients progressively, and a de-quantization module for lossy compression. Experimental results show that lossy iWave++ achieves state-of-the-art compression efficiency compared with deep network-based methods; on the Kodak dataset, lossy iWave++ leads to 17.34 percent bits saving over BPG; lossless iWave++ achieves comparable or better performance than FLIF. Our code and models are available at https://github.com/mahaichuan/Versatile-Image-Compression.

This paper presents an image compression using singular value decomposition (SVD) by extracting the red, green, and blue (RGB) channel colors. Image compression is needed in the development of various multimedia computer services and applications for example in the telecommunications and storage technologies. Now a days, video technology, digital broadcast codec and teleconferencing become popular and always requires high image compression process for display. Hence, efficient image compression is compulsory to reduce the number of storage sizes and maintain the image quality. Therefore, this article proposes image compression using SVD, which this method is efficiently reducing the image storage size and at the same time maintaining the image quality. The SVD removes redundant pixel values based on RGB colors to make the storage image size decreased. Based on the experimental analysis on two different type of image extensions (i.e., jpg and png), the SVD is capable to reduce the image size and at the same time preserving the image quality.

Objective image quality assessment (IQA) aims to predict human judgment on an image’s visual quality. This computational model is applicable in various scenarios, such as monitoring, image compression and image restoration. However, the research in this area began with natural scenes while image contents has become much more diverse. Remote computing and screen sharing applications led to the prevalence of screen content images (SCIs), which are full of texts, flat areas, and also natural scenes. Due to their different characteristics, traditional methods for IQA are insufficient for SCIs. In this paper, we show that capturing the distortions in the horizontal and vertical structures of an SCI is effective for predicting its quality. We also take into account the human visual system’s trend to analyse a scene in a multi-scale fashion. Based on these observations, we propose wavelet analysis as a means to accomplish the mentioned strategies. Experimental results on three SCI quality datasets show that plausible predictions can be made by the proposed full-reference method, while having an acceptable time complexity.

Artificial intelligence of things has brought artificial intelligence to the cutting-edge Internet of Things. In recent years, compressive sensing (CS), which relies on sparsity, is widely embedded and expected to bring more energy efficiency and a longer battery lifetime to IoT devices. Different from the other image compression standards, CS can get various reconstructed images by applying different reconstruction algorithms on coded data. Using this property, it is the first time to propose a deep learning based compressive sensing image enhancement framework using multiple reconstructed signals (CSIE-M). In this article, first, images are reconstructed by different CS reconstruction algorithms. Second, reconstructed images are assessed and sorted by a no-reference quality assessment module before being input to the quality enhancement module by order of quality scores. Finally, a multiple-input recurrent dense residual network is designed for exploiting and enriching the useful information from the reconstructed images. Experimental results show that CSIE-M obtains 1.88–8.07 dB peek-signal-to-noise (PSNR) improvement while the state-of-the-art works achieve a 1.69–6.69 dB PSNR improvement under sampling rates from 0.125 to 0.75. On the other hand, using multiple reconstructed versions of the signal can improve 0.19–0.23 dB PSNR, and only 4% reconstructing time is increasing compared to using a reconstructed signal.

In highly compressed images, i.e. quality factor $q \leq 10$ , JPEG compression causes severe compression artifacts including blocking, banding, ringing and color distortion. The compression artifacts seriously degrade image quality, which is not conducive to subsequent tasks, such as object detection and semantic segmentation. In this paper, we propose a weakly connected dense generative adversarial network for artifacts removal of highly compressed images, named WCDGAN. WCDGAN has three main ingredients of mixed convolution, weakly connected dense block (WCDB), and mixed attention. In the loss function, we add a perceptual loss to generate photo-realistic images with compression artifact removal. Experimental results show that WCDGAN successfully removes compression artifacts and produces sharp edges, clear textures and vivid colors even in highly compressed images. Moreover, WCDGAN outperforms state-of-the-art methods for compression artifact removal in terms of peak signal-to-noise ratio (PSNR) and structural similarity (SSIM).

Compactly representing visual information plays a fundamental role in optimizing the ultimate utility of myriad visual data-centered applications. Numerous approaches have been proposed to efficiently compress the texture and visual features for human visual perception and machine intelligence, respectively; however, much less work has been dedicated to studying the interactions between them. Here, we investigate the integration of feature and texture compression and show that a universal and collaborative visual information representation can be achieved in a hierarchical way. In particular, we study feature and texture compression in a scalable coding framework, where the base layer serves as the deep learning feature and the enhancement layer targets to perfectly reconstruct the texture. Based on the strong generative capability of deep neural networks, the gap between the base feature layer and enhancement layer is further filled with feature-level texture reconstruction, with the goal of further constructing texture representations from features. As such, the residuals between the original and reconstructed texture could be further conveyed in the enhancement layer. To improve the efficiency of the proposed framework, the base layer neural network is trained in a multitask manner such that the learned features enjoy both high-quality reconstruction and high-accuracy analysis. The framework and optimization strategies are further applied in face image compression, and promising coding performance has been achieved in terms of both rate-fidelity and rate-accuracy evaluations.

The amount of monitoring image information in unmanned working area of mine is large, and the hardware performance requirements are high in the image transmission and storage stage, which causes the problems of increased energy consumption and sudden decrease of the service life of sensor nodes. At present, when reconstructing mine monitoring image signal, the precision of compressive sensing measurement matrices such as Gause and Bernoulli is low. In order to solve the above problems, a new block Pascal compressive sensing measurement (BPCSM) matrix is designed. The BPCSM matrix uses the idea of non-uniform sampling and blocking in time domain, arranges multiple identical small-size Pascal matrices in a diagonal manner, and combines with the joint orthogonal matching tracking algorithm so as to realize the compression sampling and reconstruction of underground monitoring image signals. And the characteristics of orderly arrangement of row elements of Pascal matrices are used to strengthen the sampling of low frequency band of image signals so as to improve the reconstruction precision. The experimental results show that the reconstruction precision of BPCSM matrix for mine monitoring image signals is much higher than that of the commonly used measurement matrices such as Gause and Bernoulli. When the sampling rate is 0.3, the peak signal-to-noise ratio (PSNR) of the miner image reconstructed based on BPCSM matrix is about 26 dB, and the miner's facial contour is clear. When the sampling rate is 0.5, the PSNR of the miner image reconstructed based on BPCSM matrix has reached 30 dB, which can recover almost all the details of the miner image, indicating the better reconstruction performance of the BPCSM matrix. By selecting the appropriate Pascal matrix size, the reconstruction performance of the image signal can be further improved to meet the application requirements of the mine environment.

Content-weighted compression scheme for high-resolution remote-sensing (RS) images can be well modeled by Markov random field (MRF)-oriented attention. This article addresses high-resolution RS image compression by incorporating MRF into attention mechanism. To this end, we reformulate the attention mechanism with MRF-based probabilistic graph modeling implicitly and combine the target of image compression and parameter learning of MRF in a unified framework, namely high-order MRF-oriented attention (HMA) network. Specifically, HMA extends key-value query (KVQ) pairwise terms of the vanilla attention to high-order terms, by which the prior information could be expressed effectively to boost performance of high-resolution RS image compression. It is noted that several superiorities of HMA are listed. First, unlike the vanilla attention network that apt to yield coarse features, HMA is capable of output more pleasing decoding results. Second, HMA can accelerate the convergence in the training of the deep neural networks (DNNs), thus facilitating deploying it on resource-limited IOT devices. Third, HMA demonstrates its potential of processing semantic joint task. Moreover, We thoroughly evaluate our approach on standard data sets of varying resolutions, the proposed framework performs favorably against most image coding standards and DNN-based codecs on the ISPRS Vaihingen data set and the USC-SIPI data set especially at low bit rates.

Earth observation through satellite images is crucial to help economic activities as well as to monitor the impact of human activities on ecosystems. Current satellite systems are subjected to strong computational complexity constraints. Thus, image compression is perfomed onboard with specifically tailored algorithms while image denoising is performed on the ground. In this letter, we intend to address satellite image compression and denoising with neural networks. The first proposed approach uses a single neural architecture for joint onboard compression and denoising. The second proposed approach sequentially uses a first neural architecture for onboard compression and a second one for on ground denoising. For both approaches, the onboard architectures are lightened as much as possible, following the procedure proposed in [1]. The two approaches are shown to outperform the current satellite imaging system and their respective pros and cons are discussed.

Recently, learning-based image compression model has attracted much attention due to its impressive performance and ease of optimization, compared with traditional DCT and wavelet-based image compression standards. Most learning-based image compression models are trained to minimize joint rate-distortion (RD) loss on one single RD trade-off point. However, in many multimedia applications, due to communication constraints, or display adaptation needs for different spatial formats, bit rates or power, it is necessary to provide a variety of image versions for different client devices. To fulfill this requirement, typical end-to-end image compression methods have to compress an image into several bit streams independently by a number of pre-trained networks, which are resource-consuming because of redundancy among these streams. To address this problem, inspired by traditional scalable video coding framework, we propose a learning-based end-to-end quality and spatial scalable image compression (QSSIC) model in multi-layer structure, in which each layer could generate one bitstream corresponding to a specified resolution and image fidelity. This scalability is achieved by exploring the potential of feature-domain representation prediction and reuse. To be specific, firstly, bitstreams of previous layers are used to predict the current layer representations which contains the enhancement information, and then only prediction residuals need to be coded in enhancement layers. Secondly, previous bitstreams are reused in image reconstruction in higher layers to provide basic information. The proposed model could be optimized in an end-to-end manner. Extensive experiments show that our method outperforms state-of-art deep neural networks (DNN)-based auto-encoders in simulcast scenarios. In addition, our method has a better performance than the traditional scalable image compression method scalable extension of H.264/AVC (SVC) and is comparable to scalable extension of H.265/HEVC (SHVC).

Given the uniqueness of synthetic aperture radar (SAR) images, traditional optical image compression algorithms cannot fully exploit their redundant information. To improve SAR image compression in terms of rate-distortion performance and visual perception, an end-to-end SAR image compression convolutional neural network (CNN) model based on a variational autoencoder is proposed. The proposed CNN model consists of a main autoencoder and a hyper autoencoder. To reduce dependencies in latent space, a joint transform of linear CNN and nonlinear generalized divisive normalization (GDN) activation is applied in the main autoencoder. Moreover, residual blocks are combined with the transforms to boost the efficiency of feature learning and make use of subpixels to improve the quality of reconstructed images. Instead of a fixed entropy model, a conditioned entropy model that works with a hyperprior network is used to learn the distribution of latents, which helps to further improve the compression quality. During training, the model is optimized by evaluating the rate-distortion performance. The experimental results show that the proposed method can achieve better distortion performance than JPEG, JPEG2000, and the available CNN-based method in terms of objective evaluation criteria and human vision perception quality.

Underwater wireless sensor network (UWSN) comprises of large number of sensors and underwater vehicles deployed collaboratively to perform data collection, interpretation and processing. These sensors and vehicles have been equipped with cameras to capture a visual picture of underwater targets and precious resources in recent times. These cameras in the sensors generate large volumes of data as they are continuously involved in surveillance of the aquatic environment. Transmitting or storing the data as a whole drains the power of the battery-operated nodes. Therefore it is necessary to reduce the data to save energy and improve the lifetime of the sensors. To serve this purpose, block compressed sensing (BCS) based image compression can be used. However, BCS has two significant issues: low sampling efficiency for poorly sparsed real-time underwater images and fixing the samples chosen from image blocks. The former can be overcome by permuting and evenly distributing the transform coefficients to all blocks. Similarly, the latter can be overcome by adopting adaptive block compressed sensing (ABCS). A combination of coefficient permutation and ABCS, namely coefficient permuted adaptive block compressed sensing (CP-ABCS), is proposed for better image reconstruction with fewer samples. This proposed approach operates in the processor of the sensors to compress data within the nodes and then transmit the reduced data. It has improved PSNR of 4-8dB, SSIM of 0.1-0.3 and space-saving (SS) of 5-10% compared with other literature schemes. Also, it has used only significantly fewer samples of about 10-20% for reconstruction.

Discrete transforms are used in many signal processing applications such as audio compression, image compression, video compression, high efficiency video coding (HEVC), and so on. This paper proposes a generic VLSI architecture for performing ( N × N × N )-point discrete transformation. This proposed generic architecture can be used to perform the 3D discrete transformations such as Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), Discrete Hartley Transform, Integer Discrete Cosine Transform (Integer DCT), Discrete Hadamard Transform, and Discrete Walsh Transform. Also, this ( N × N × N )-point discrete transformation’s proposed architecture is to perform ⌊ N 2 i ⌋ numbers of ( 2 i × 2 i × 2 i )-point discrete transforms in parallel, where i is varied from 2, 3, 4, … l o g 2 N for Discrete Hartley Transform, i = 3 for DCT/DST, and i = 2 , 3 , 4 , 5 , 6 for Integer DCT/Discrete Walsh/Discrete Hadamard Transforms. Also, our proposed architecture eliminates the requirement of larger storage buffer in between the row, column, and temporal processes. The trade-off in our proposed design is the number of cycles to complete the operation. All the existing and proposed techniques are implemented with 45 nm CMOS technology using Cadence. The synthesis results show that our proposed architecture achieves 73% of reduction in critical path delay as compared with the parallel Butterfly architecture based 3D-DCT.

International audience; Lifting-based wavelet transform has been extensively used for efficient compression of various types of visual data. Generally, the performance of such coding schemes strongly depends on the lifting operators used, namely the prediction and update filters. Unlike conventional schemes based on linear filters, we propose, in this paper, to learn these operators by exploiting neural networks. More precisely, a classical Fully Connected Neural Network (FCNN) architecture is firstly employed to perform the prediction and update. Then, we propose to improve this FCNN-based Lifting Scheme (LS) in order to better take into account the input image to be encoded. Thus, a novel dynamical FCNN model is developed, making the learning process adaptive to the input image contents for which two adaptive learning techniques are proposed. While the first one resorts to an iterative algorithm where the computation of two kinds of variables is performed in an alternating manner, the second learning method aims to learn the model parameters directly through a reformulation of the loss function. Experimental results carried out on various test images show the benefits of the proposed approaches in the context of lossy and lossless image compression.

Satellite image compression technology plays an important role in the development of space science. As optical sensors on satellites become more sophisticated, high-resolution and high-fidelity satellite images will occupy more storage. This raises the required transmission bandwidth and transmission rate in the satellite–ground data transmission system. In order to reduce the pressure from image transmission on the data transmission system, a spaceborne target detection system based on Yolov5 and a satellite image compression transmission system is proposed in this paper. It can reduce the pressure on the data transmission system by detecting the object of interest and deciding whether to transmit. An improved Yolov5 network is proposed to detect the small target on the high-resolution satellite image. Simulation results show that the improved Yolov5 network proposed in this paper can detect specific targets in real satellite images, including aircraft, ships, etc. At the same time, image compression has little effect on target detection, so detection complexity can be effectively reduced and detection speed can be improved by detecting the compressed images.

End-to-end learned image and video codecs, based on auto-encoder architecture, adapt naturally to image resolution, thanks to their convolutional aspect. However, while coding high resolution images, these codecs face hardware problems such as memory saturation. This paper proposes a patch-based image coding solution based on an end-to-end learned model, which aims to remedy to the hardware limitation while maintaining the same quality as full resolution image coding. Our method consists in coding overlapping patches of the image and reconstructing them into a decoded image using a weighting function. This approach manages to be on par with the performance of full resolution image coding using an endto-end learned model, and even slightly outperforms it, while being adaptable to different memory sizes. Moreover, this work undertakes a full study on the effect of the patch size on this solution’s performance, and consequently determines the best patch resolution in terms of coding time and coding efficiency. Finally, the method introduced in this work is also compatible with any learned codec based on a conv/deconvolutional autoencoder architecture without having to retrain the model.&nbsp

Image compression research has increased significantly because of increasing demands for image transmission in computer and mobile environments. When calculating the amount of bits per image from conventional quantization methods and sampling rates, image compression is required. As a result, it has become necessary to develop efficient image compression techniques. Color images are in trend these days during communication. Most of the researchers have worked only on grayscale image compression. Colored image channels have been handled exclusively for many years, or color image is transformed into grayscale image. Improvements have been made to image compression algorithms SVD and DCT. In this paper, seven standard images have been used for compression using DCT and SVD, individually for experimental purposes. The performance of the model is measure on the basis of various performance matrices like Peak Signal to Noise Ratio, Mean Square Error, Normalized Co-relation (NC), Volumetric Efficiency, and percent space savings.

The non-deterministic characteristics of the floating coding scheme for the method of cryptocompression representation of images in a differentiated basis are estimated. Namely: eestimation of the number of image elements that form code constructions; estimation of the length of the formed code structures. It is proved that code constructions are formed on a variable (predetermined) number of elements of the original image. In the formation of thecode of the information component of the cryptocompression representation of images in a differentiated basis with acodeword length of 64 bitsmay involve from 8 to 64 or more elements of the original image. Code constructs areformed of variable (predetermined) length, which ranges from 57 to 64 bits with a codeword length of 64 bits Withoutan open system of grounds, it is impossible to priori predict the length of any code of the informational component ofthe cryptocompression image representation. Moreover, it is impossible to break all information component into separate blocks corresponding to separate code constructions. The number of elements that form the codes of the information component of the cryptocompression representation of images, and the lengths of the information componentsthemselves depend only on the initial values of the image elements. They are different for different images and for different color planes within the same image Проведено оцінку недетермінованих характеристик плаваючої схеми кодування для методу криптокомпресійного представлення зображень в диференційованому базисі. А саме: оцінка кількості елементів зображення, які формують кодові конструкції; оцінка довжини сформованих кодових конструкцій. Доведено, що кодові конструкції формуються на змінної (заздалегідь невизначеної) кількості елементів вихідного зображення. У формуванні коду інформаційної складової криптокомпресійного представлення зображень в диференційованому базисі при довжині кодового слова в 64 біта може брати участь від 8 до 64 і більше елементів вихідного зображення. Кодові конструкції формуються змінної (заздалегідь невизначеної) довжини, яка знаходиться в діапазоні від 57 до 64 біт при довжині кодового слова в 64 біта. Без наявності відкритої системи підстав неможливо апріорно передбачити довжину будь-якого коду інформаційної складової криптокомпресійного представлення зображень. Тим більше, не можна розбити всю інформаційну складову на окремі блоки, що відповідають окремим кодовим конструкціям. Кількість елементів, що формують коди інформаційної складової криптокомпресійного представлення зображень, і довжини самих інформаційних складових залежать тільки від вихідних значень елементів зображення. Вони є різними, як для різних зображень, так і для різних колірних площин в межах одного зображення.

It has always been a major issue for a hospital to acquire real-time information about a patient in emergency situations. Because of this, this research presents a novel high-compression-ratio and real-time-process image compression very-large-scale integration (VLSI) design for image sensors in the Internet of Things (IoT). The design consists of a YEF transform, color sampling, block truncation coding (BTC), threshold optimization, sub-sampling, prediction, quantization, and Golomb–Rice coding. By using machine learning, different BTC parameters are trained to achieve the optimal solution given the parameters. Two optimal reconstruction values and bitmaps for each 4 × 4 block are achieved. An image is divided into 4 × 4 blocks by BTC for numerical conversion and removing inter-pixel redundancy. The sub-sampling, prediction, and quantization steps are performed to reduce redundant information. Finally, the value with a high probability will be coded using Golomb–Rice coding. The proposed algorithm has a higher compression ratio than traditional BTC-based image compression algorithms. Moreover, this research also proposes a real-time image compression chip design based on low-complexity and pipelined architecture by using TSMC 0.18 μm CMOS technology. The operating frequency of the chip can achieve 100 MHz. The core area and the number of logic gates are 598,880 μm2 and 56.3 K, respectively. In addition, this design achieves 50 frames per second, which is suitable for real-time CMOS image sensor compression.

We face many challenges in deploying high-performance neural networks in practice. These challenges are predominantly due to the size of neural networks and apply to both training and inference. Compressing neural networks to make them train and run more efficiently is therefore crucial and has been a parallel line of research from the early days of neural networks development. The two main compression techniques in deep learning, which are the focus of this thesis, are pruning and quantization. This thesis explores how the information from higher-order gradients (meta-gradients) be used to improve deep learning compression. We start by identifying a fundamental limitation in the formulation of pruning: Although many methods, such as saliency-based pruning, follow pruning by a training or fine-tuning stage, parameter saliencies only look at a snapshot of parameters without taking into account the "trainability" of the parameters. We show how meta-gradients can be used as a more informative signal to find better trainable subnetworks at initialization. We then look at quantized neural networks and show how meta-gradients can be used in a regularization scheme to "learn" models with inherent robustness against post-training quantization. Finally, we look at the dual compression problem, i.e. using neural networks to compress data sources. We start with images and propose a simple autoencoder-free architecture where we store weights of a neural network instead of RGB values of image pixels. We then use meta-gradients to meta-learn a base network to amortize the cost of training one network per input. A significant advantage of our learning compression is that it becomes agnostic to the data type, and we present results on various data types beyond 2D images. Importantly, we evaluate the usefulness of standard DNN compression techniques, e.g., quantization, for this new type of neural network.

In this study, we proposed a region of interest (ROI) compression algorithm under the deep learning self-encoder framework to improve the reconstruction performance of the image and reduce the distortion of the ROI. First, we adopted a remote sensing image cloud detection algorithm for detecting important targets in images, that is, separating the remote sensing background from important regions in remote sensing images and then determining the target regions because most traditional ROI-based image compression algorithms utilize the manual labeling of the ROI to achieve region separation in images. We designed a multiscale ROI self-coding network from coarse to fine with a hierarchical super priority layer to synthesize images to reduce the spatial redundancy more effectively, thus greatly improving the distortion rate performance of image compression. By using a spatial attention mechanism for the ROI in the image compression network, we achieved better compression performance.

Small satellites empower different applications for an affordable price. By dealing with a limited capacity for using instruments with high power consumption or high data-rate requirements, small satellite missions usually focus on specific monitoring and observation tasks. Considering that multispectral and hyperspectral sensors generate a significant amount of data subjected to communication channel impairments, bandwidth constraint is an important challenge in data transmission. That issue is addressed mainly by source and channel coding techniques aiming at an effective transmission. This paper targets a significant further bandwidth reduction by proposing an on-the-fly analysis on the satellite to decide which information is effectively useful before coding and transmitting. The images are tiled and classified using a set of detection algorithms after defining the least relevant content for general remote sensing applications. The methodology makes use of the red-band, green-band, blue-band, and near-infrared-band measurements to perform the classification of the content by managing a cloud detection algorithm, a change detection algorithm, and a vessel detection algorithm. Experiments for a set of typical scenarios of summer and winter days in Stockholm, Sweden, were conducted, and the results show that non-important content can be identified and discarded without compromising the predefined useful information for water and dry-land regions. For the evaluated images, only 22.3% of the information would need to be transmitted to the ground station to ensure the acquisition of all the important content, which illustrates the merits of the proposed method. Furthermore, the embedded platform’s constraints regarding processing time were analyzed by running the detection algorithms on Unibap’s iX10-100 space cloud platform.

Learning objectives Background Findings and procedure details Conclusion Personal information and conflict of interest References, Learning objectives: Evaluate tailgut cysts (retrorectal cystic hamartoma) imaging findings, main differential diagnosis, clinical presentation, and...

Learning objectives Background Findings and procedure details Conclusion Personal information and conflict of interest References, Learning objectives: This paper focuses not only on exposing the main ultrasound artifacts, but also on how to turn them into our allies when interpreting B-mode abdominal ultrasound, in order to make this imaging technique an easier tool to use without fear of making...

Purpose Methods and materials Results Conclusion Personal information and conflict of interest References, Purpose: The clivus is a bony structure at the base of the skull. It is a shallow depression situated behind the sellar dorsum of the sphenoid bone. It slopes gradually to the anterior portion of the basilar occipital bone at the sphenoid bone's...

Purpose Methods and materials Results Conclusion Personal information and conflict of interest References, Purpose: Despite increased radiation exposure, computer tomography (CT) is a widely used diagnostic tool in acute life-threatening situations and also in monitoring of chronic, long-standing thoracic diseases. In pediatric patients who are more sensitive to...

Sparse Voxel Directed Acyclic Graphs (SVDAGs) are an efficient solution for storing high-resolution voxel geometry. Recently, algorithms for the interactive modification of SVDAGs have been proposed that maintain the compressed geometric representation. Nevertheless, voxel attributes, such as colours, require an uncompressed storage, which can result in high memory usage over the course of the application. The reason is the high cost of existing attribute-compression schemes which remain unfit for interactive applications. In this paper, we introduce two attribute compression methods (lossless and lossy), which enable the interactive editing of compressed high-resolution voxel scenes including attributes.

Learning objectives Background Findings and procedure details Conclusion Personal information and conflict of interest References, Learning objectives: We aim to describe radiological findings of peritoneal tuberculosis in CT and to show the keys of the main differential...

Learning objectives Background Findings and procedure details Conclusion Personal information and conflict of interest References, Learning objectives: During the last world pandemic, the medical education turned into a challenging task. The setting of the classes we used to managed, has entirely changed. Ironically, during COVID-19, the rad residents were in charge and proposed an active learning...

Purpose Methods and materials Results Conclusion Personal information and conflict of interest References, Purpose: To determine the optimal energy reconstruction for evaluating rectal tumor and to compare the diagnostic performance related-depiction grade of second-generation dual-layer spectral CT and conventional polychromatic CT with MRI results as reference...

El trabajo consiste en tres partes diferentes pero relacionadas entre sí. Primero se profundiza el estudio de la compresibilidad de datos 1D y 2D. Esto incluye entender qué es la compresibilidad y por qué la entropía de Shannon no sirve a nuestros objetivos. Segundo, se define una métrica de compresibilidad para datos 2D. También se implementan dos métricas de compresibilidad para datos 1D ya existentes (Lempel-Ziv y Transformada de Burrows-Wheeler). Por último, se realizan experimentos para comparar el rendimiento de las métricas con diferentes conjuntos de imágenes y comprobar si la nueva métrica está bien definida o captura bien la compresibilidad de las imágenes. The work has been divided into three different but related parts. Firstly, an in-depth study of 1D and 2D data compressibility is carried out. This includes understanding what is compressibility and why Shannon's entropy is not useful for our purposes. Secondly, a 2D data compressibility metric is defined. Also, two existing 1D data compressibility metrics are implemented (Lempel-Ziv and Burrows-Wheeler transform). Finally, a series of experiments are run in order to compare the metrics perfomances with several sets of images and to test whether the new metric is well defined and is able to capture the compressibility of the images. El treball consisteix en tres parts diferents però relacionades entre si. Primer s'aprofundeix l'estudi de la compressibilitat de dades 1D i 2D. Això inclou entendre què és la compressibilitat i per què l'entropia de Shannon no serveix als nostres objectius. Segon, es defineix una mètrica de compressibilitat per a dades 2D. També s'implementen dues mètriques de compressibilitat per a dades 1D ja existents (Lempel-Ziv i Transformada de Burrows-Wheeler). Finalment, es realitzen experiments per a comparar el rendiment de les mètriques amb diferents conjunts d'imatges i comprovar si la nova mètrica està ben definida o captura bé la compressibilitat de les imatges.

Purpose Methods and materials Results Conclusion Personal information and conflict of interest References, Purpose: Review the anatomy of the hamstring and the rectus femoral muscles. To review the imaging findings and classify hamstring and rectus femoral injuries. Review the mechanism of injury, complications and recovery times in the non-professional...

Purpose Methods and materials Results Conclusion Personal information and conflict of interest References, Purpose: The clear adaptive low-noise method (CaLM) based on the non-local-mean method can possibly improve the image quality and quantitative performance of PET by reducing the loss of lesion contrast caused by smoothing. The current study aimed to...

Learning objectives Background Findings and procedure details Conclusion Personal information and conflict of interest References, Learning objectives: To review the most common incidental cardiac findings on routine chest-CT examinations. To recognize which cardiac findings are relevant and can alter the clinical course and outcome of...

Learning objectives Background Findings and procedure details Conclusion Personal information and conflict of interest References, Learning objectives: The relevance of venilation / perfusion (V/Q) scintigraphy in the diagnosis of pulmonary embolism (PE) are well-accepted and frequently performed procedure in the diagnosis of pulmonary embolism, there has dwindled in recent years. Despite the fact...

Purpose Methods and materials Results Conclusion Personal information and conflict of interest References, Purpose: In Japan, cerebrovascular disease is the fourth cause of death and the first cause of besridden patients 1) . Recently, cerebral stroke has been under serious consideration due to the westernization of the diets and an increase in geriatric...

This paper reviews the Challenge on Super-Resolution of Compressed Image and Video at AIM 2022. This challenge includes two tracks. Track 1 aims at the super-resolution of compressed image, and Track 2 targets the super-resolution of compressed video. In Track 1, we use the popular dataset DIV2K as the training, validation and test sets. In Track 2, we propose the LDV 3.0 dataset, which contains 365 videos, including the LDV 2.0 dataset (335 videos) and 30 additional videos. In this challenge, there are 12 teams and 2 teams that submitted the final results to Track 1 and Track 2, respectively. The proposed methods and solutions gauge the state-of-the-art of super-resolution on compressed image and video. The proposed LDV 3.0 dataset is available at https://github.com/RenYang-home/LDV_dataset. The homepage of this challenge is at https://github.com/RenYang-home/AIM22_CompressSR.

Treballs Finals de Grau de Física, Facultat de Física, Universitat de Barcelona, Curs: 2022-2023, Tutors: Alberto Fernández-Nieves, Ramon Planet, We use file compression techniques to characterize colloidal gelation. The colloidal system is built through a diffusion-limited aggregation simulation, and at the gelation point, the growing dynamics are changed to emulate the transition. We compress images of the system before and after the gelation point. Using the computable information density (CID) we have been able to ascribe a second-order character to the transition.