Showing papers on "JPEG 2000 published in 2017"

Lossy Image Compression with Compressive Autoencoders

Abstract: We propose a new approach to the problem of optimizing autoencoders for lossy image compression. New media formats, changing hardware technology, as well as diverse requirements and content types create a need for compression algorithms which are more flexible than existing codecs. Autoencoders have the potential to address this need, but are difficult to optimize directly due to the inherent non-differentiabilty of the compression loss. We here show that minimal changes to the loss are sufficient to train deep autoencoders competitive with JPEG 2000 and outperforming recently proposed approaches based on RNNs. Our network is furthermore computationally efficient thanks to a sub-pixel architecture, which makes it suitable for high-resolution images. This is in contrast to previous work on autoencoders for compression using coarser approximations, shallower architectures, computationally expensive methods, or focusing on small images.

Lossy Image Compression with Compressive Autoencoders

Abstract: We propose a new approach to the problem of optimizing autoencoders for lossy image compression. New media formats, changing hardware technology, as well as diverse requirements and content types create a need for compression algorithms which are more flexible than existing codecs. Autoencoders have the potential to address this need, but are difficult to optimize directly due to the inherent non-differentiabilty of the compression loss. We here show that minimal changes to the loss are sufficient to train deep autoencoders competitive with JPEG 2000 and outperforming recently proposed approaches based on RNNs. Our network is furthermore computationally efficient thanks to a sub-pixel architecture, which makes it suitable for high-resolution images. This is in contrast to previous work on autoencoders for compression using coarser approximations, shallower architectures, computationally expensive methods, or focusing on small images.

Real-time adaptive image compression

Abstract: We present a machine learning-based approach to lossy image compression which outperforms all existing codecs, while running in real-time. Our algorithm typically produces files 2.5 times smaller than JPEG and JPEG 2000, 2 times smaller than WebP, and 1.7 times smaller than BPG on datasets of generic images across all quality levels. At the same time, our codec is designed to be lightweight and deployable: for example, it can encode or decode the Kodak dataset in around 10ms per image on GPU. Our architecture is an autoencoder featuring pyramidal analysis, an adaptive coding module, and regularization of the expected codelength. We also supplement our approach with adversarial training specialized towards use in a compression setting: this enables us to produce visually pleasing reconstructions for very low bitrates.

Real-Time Adaptive Image Compression

Abstract: We present a machine learning-based approach to lossy image compression which outperforms all existing codecs, while running in real-time Our algorithm typically produces files 25 times smaller than JPEG and JPEG 2000, 2 times smaller than WebP, and 17 times smaller than BPG on datasets of generic images across all quality levels At the same time, our codec is designed to be lightweight and deployable: for example, it can encode or decode the Kodak dataset in around 10ms per image on GPU Our architecture is an autoencoder featuring pyramidal analysis, an adaptive coding module, and regularization of the expected codelength We also supplement our approach with adversarial training specialized towards use in a compression setting: this enables us to produce visually pleasing reconstructions for very low bitrates

Learning Convolutional Networks for Content-weighted Image Compression

Abstract: Lossy image compression is generally formulated as a joint rate-distortion optimization to learn encoder, quantizer, and decoder. However, the quantizer is non-differentiable, and discrete entropy estimation usually is required for rate control. These make it very challenging to develop a convolutional network (CNN)-based image compression system. In this paper, motivated by that the local information content is spatially variant in an image, we suggest that the bit rate of the different parts of the image should be adapted to local content. And the content aware bit rate is allocated under the guidance of a content-weighted importance map. Thus, the sum of the importance map can serve as a continuous alternative of discrete entropy estimation to control compression rate. And binarizer is adopted to quantize the output of encoder due to the binarization scheme is also directly defined by the importance map. Furthermore, a proxy function is introduced for binary operation in backward propagation to make it differentiable. Therefore, the encoder, decoder, binarizer and importance map can be jointly optimized in an end-to-end manner by using a subset of the ImageNet database. In low bit rate image compression, experiments show that our system significantly outperforms JPEG and JPEG 2000 by structural similarity (SSIM) index, and can produce the much better visual result with sharp edges, rich textures, and fewer artifacts.

Improved Lossy Image Compression with Priming and Spatially Adaptive Bit Rates for Recurrent Networks

Abstract: We propose a method for lossy image compression based on recurrent, convolutional neural networks that outperforms BPG (4:2:0 ), WebP, JPEG2000, and JPEG as measured by MS-SSIM. We introduce three improvements over previous research that lead to this state-of-the-art result. First, we show that training with a pixel-wise loss weighted by SSIM increases reconstruction quality according to several metrics. Second, we modify the recurrent architecture to improve spatial diffusion, which allows the network to more effectively capture and propagate image information through the network's hidden state. Finally, in addition to lossless entropy coding, we use a spatially adaptive bit allocation algorithm to more efficiently use the limited number of bits to encode visually complex image regions. We evaluate our method on the Kodak and Tecnick image sets and compare against standard codecs as well recently published methods based on deep neural networks.

Lossless Compression of Medical Images Using 3-D Predictors

Abstract: This paper describes a highly efficient method for lossless compression of volumetric sets of medical images, such as CTs or MRIs. The proposed method, referred to as 3-D-MRP, is based on the principle of minimum rate predictors (MRPs), which is one of the state-of-the-art lossless compression technologies presented in the data compression literature. The main features of the proposed method include the use of 3-D predictors, 3-D-block octree partitioning and classification, volume-based optimization, and support for 16-b-depth images. Experimental results demonstrate the efficiency of the 3-D-MRP algorithm for the compression of volumetric sets of medical images, achieving gains above 15% and 12% for 8- and 16-bit-depth contents, respectively, when compared with JPEG-LS, JPEG2000, CALIC, and HEVC, as well as other proposals based on the MRP algorithm.

An End-to-End Compression Framework Based on Convolutional Neural Networks

Abstract: Deep learning, e.g., convolutional neural networks (CNNs), has achieved great success in image processing and computer vision especially in high level vision applications such as recognition and understanding. However, it is rarely used to solve low-level vision problems such as image compression studied in this paper. Here, we move forward a step and propose a novel compression framework based on CNNs. To achieve high-quality image compression at low bit rates, two CNNs are seamlessly integrated into an end-to-end compression framework. The first CNN, named compact convolutional neural network (ComCNN), learns an optimal compact representation from an input image, which preserves the structural information and is then encoded using an image codec (e.g., JPEG, JPEG2000 or BPG). The second CNN, named reconstruction convolutional neural network (RecCNN), is used to reconstruct the decoded image with high-quality in the decoding end. To make two CNNs effectively collaborate, we develop a unified end-to-end learning algorithm to simultaneously learn ComCNN and RecCNN, which facilitates the accurate reconstruction of the decoded image using RecCNN. Such a design also makes the proposed compression framework compatible with existing image coding standards. Experimental results validate that the proposed compression framework greatly outperforms several compression frameworks that use existing image coding standards with state-of-the-art deblocking or denoising post-processing methods.

Semantic Perceptual Image Compression Using Deep Convolution Networks

Abstract: It has long been considered a significant problem to improve the visual quality of lossy imageand video compression. Recent advances in computing power together with the availabilityof large training data sets has increased interest in the application of deep learning cnnsto address image recognition and image processing tasks. Here, we present a powerful cnntailored to the specific task of semantic image understanding to achieve higher visual qualityin lossy compression. A modest increase in complexity is incorporated to the encoder whichallows a standard, off-the-shelf jpeg decoder to be used. While jpeg encoding may beoptimized for generic images, the process is ultimately unaware of the specific content ofthe image to be compressed. Our technique makes jpeg content-aware by designing andtraining a model to identify multiple semantic regions in a given image. Unlike objectdetection techniques, our model does not require labeling of object positions and is able toidentify objects in a single pass. We present a new cnn architecture directed specifically toimage compression, which generates a map that highlights semantically-salient regions sothat they can be encoded at higher quality as compared to background regions. By addinga complete set of features for every class, and then taking a threshold over the sum of allfeature activations, we generate a map that highlights semantically-salient regions so thatthey can be encoded at a better quality compared to background regions. Experimentsare presented on the Kodak PhotoCD dataset and the MIT Saliency Benchmark dataset, in which our algorithm achieves higher visual quality for the same compressed size whilepreserving PSNR1.

Lossy compression of lenslet images from plenoptic cameras combining sparse predictive coding and JPEG 2000

Abstract: This paper proposes a lenslet image compression method scalable from low bitrates to fully lossless. The subaperture images are split into two sets: a set of reference views, encoded by a standard lossy or lossless compressor, and the set of dependent views, which are reconstructed by sparse prediction from the reference set using the geometrical information from the depth map. The set of reference views may contain all views and all views may also be dependent views, in which case the sparse predictive stage does not reconstruct from scratch the views, but it refines in a sequential order all views by combining in an optimal way the information about the same region existing in neighbor views. The encoder transmits to the decoder a segmented version of the scene depthmap, the encoded versions of the reference views, displacements for each region from the central view to each of the dependent views, and finally the sparse predictors for each region and each dependent view. The scheme can be configured to ensure random access to the dependent views, while the reference views are compressed in a backward compatible way, e.g., using JPEG 2000. The experimental results show performance better than that of the baseline standard compressor used, JPEG 2000.

JPEG at 25: Still Going Strong

Abstract: JPEG is celebrating the 25th anniversary of its approval as a standard this year. Where did JPEG come from, and what are the fundamental components that have given it longevity?

An End-to-End Compression Framework Based on Convolutional Neural Networks

Abstract: Traditional image coding standards (such as JPEG and JPEG2000) make the decoded image suffer from many blocking artifacts or noises since the use of big quantization steps. To overcome this problem, we proposed an end-to-end compression framework based on two CNNs, as shown in Figure 1, which produce a compact representation for encoding using a third party coding standard and reconstruct the decoded image, respectively. To make two CNNs effectively collaborate, we develop a unified end-to-end learning framework to simultaneously learn CrCNN and ReCNN such that the compact representation obtained by CrCNN preserves the structural information of the image, which facilitates to accurately reconstruct the decoded image using ReCNN and also makes the proposed compression framework compatible with existing image coding standards.

Predictive Lossless Compression of Regions of Interest in Hyperspectral Images With No-Data Regions

Abstract: This paper addresses the problem of efficient predictive lossless compression on the regions of interest (ROIs) in the hyperspectral images with no-data regions. We propose a two-stage prediction scheme, where a context-similarity-based weighted average prediction is followed by recursive least square filtering to decorrelate the hyperspectral images for compression. We then propose to apply separate Golomb–Rice codes for coding the prediction residuals of the full-context pixels and boundary pixels, respectively. To study the coding gains of this separate coding scheme, we introduce a mixture geometric model to represent the residuals associated with various combinations of the full-context pixels and boundary pixels. Both information-theoretic analysis and simulations on synthetic data confirm the advantage of the separate coding scheme over the conventional coding method based on a single underlying geometric distribution. We apply the aforementioned prediction and coding methods to four publicly available hyperspectral image data sets, attaining significant improvements over several other state-of-the-art methods, including the shape-adaptive JPEG 2000 method.

No-Reference JPEG Image Quality Assessment Based on Blockiness and Luminance Change

Abstract: When scoring the quality of JPEG images, the two main considerations for viewers are blocking artifacts and improper luminance changes, such as blur. In this letter, we first propose two measures to estimate the blockiness and the luminance change within individual blocks. Then, a no-reference image quality assessment (NR-IQA) method for JPEG images is proposed. Our method obtains the quality score by considering the blocking artifacts and the luminance changes from all nonoverlapping 8 × 8 blocks in one JPEG image. The proposed method has been tested on five public IQA databases and compared with five state-of-the-art NR-IQA methods for JPEG images. The experimental results show that our method is more consistent with subjective evaluations than the state-of-the-art NR-IQA methods. The MATLAB source code of our method is available at http://image.ustc.edu.cn/IQA.html .

JPEG 2000 compression of unfocused light field images based on lenslet array slicing

Abstract: This paper proposes an algorithm for lossy compression of unfocused light field images. The raw light field is preprocessed by demosaicing, devignetting and slicing of the raw lenset array image. The slices are then rearranged in tiles and compressed by the standard JPEG 2000 encoder. The experimental analysis compares the performance of the proposed method against the direct compression with JPEG 2000, and JPEG XR, in terms of BD-PSNR gain and bit rate reduction. Obtained results show that the proposed method outperforms direct application of the reference state of the art image encoders.

Guetzli: Perceptually Guided JPEG Encoder

Abstract: Guetzli is a new JPEG encoder that aims to produce visually indistinguishable images at a lower bit-rate than other common JPEG encoders. It optimizes both the JPEG global quantization tables and the DCT coefficient values in each JPEG block using a closed-loop optimizer. Guetzli uses Butteraugli, our perceptual distance metric, as the source of feedback in its optimization process. We reach a 29-45% reduction in data size for a given perceptual distance, according to Butteraugli, in comparison to other compressors we tried. Guetzli's computation is currently extremely slow, which limits its applicability to compressing static content and serving as a proof- of-concept that we can achieve significant reductions in size by combining advanced psychovisual models with lossy compression techniques.

JPEG XS, a new standard for visually lossless low-latency lightweight image compression

Abstract: JPEG XS is an upcoming standard from the JPEG Committee (formally known as ISO/IEC SC29 WG1). It aims to provide an interoperable visually lossless low-latency lightweight codec for a wide range of applications including mezzanine compression in broadcast and Pro-AV markets. This requires optimal support of a wide range of implementation technologies such as FPGAs, CPUs and GPUs. Targeted use cases are professional video links, IP transport, Ethernet transport, real-time video storage, video memory buffers, and omnidirectional video capture and rendering. In addition to the evaluation of the visual transparency of the selected technologies, a detailed analysis of the hardware and software complexity as well as the latency has been done to make sure that the new codec meets the requirements of the above-mentioned use cases. In particular, the end-to-end latency has been constrained to a maximum of 32 lines. Concerning the hardware complexity, neither encoder nor decoder should require more than 50% of an FPGA similar to Xilinx Artix 7 or 25% of an FPGA similar to Altera Cyclon 5. This process resulted in a coding scheme made of an optional color transform, a wavelet transform, the entropy coding of the highest magnitude level of groups of coefficients, and the raw inclusion of the truncated wavelet coefficients. This paper presents the details and status of the standardization process, a technical description of the future standard, and the latest performance evaluation results.

Lossless compression of color filter array mosaic images with visualization via JPEG 2000

Abstract: Digital cameras have become ubiquitous for amateur and professional applications. The raw images captured by digital sensors typically take the form of color filter array (CFA) mosaic images, which must be "developed" (via digital signal processing) before they can be viewed. Photographers and scientists often repeat the "development process" using different parameters to obtain images suitable for different purposes. Since the development process is generally not invertible, it is commonly desirable to store the raw (or undeveloped) mosaic images indefinitely. Uncompressed mosaic image file sizes can be more than 30 times larger than those of developed images stored in JPEG format. Thus, data compression is of interest. Several compression methods for mosaic images have been proposed in the literature. However, they all require a custom decompressor followed by development-specific software to generate a displayable image. In this paper, a novel compression pipeline that removes these requirements is proposed. Specifically, mosaic images can be losslessly recovered from the resulting compressed files, and, more significantly, images can be directly viewed (decompressed and developed) using only a JPEG 2000 compliant image viewer. Experiments reveal that the proposed pipeline attains excellent visual quality, while providing compression performance competitive to that of state-of-the-art compression algorithms for mosaic images.

Detection of operation chain

Abstract: The goal of forensic investigators is to reveal the processing history of a digital image. Many forensic techniques are devoted to detecting the intrinsic traces left by image processing and tampering. However, existing forensic techniques are easily defeated in presence of pre- and post-processing. In real scenarios, images may be sequentially manipulated by a series of operations (the so called operation chain). This paper addresses the operation chain consisting of JPEG compression and resampling. The transformed block artifacts (TBAG) characterizing this operation chain are analysed at both the pixel and discrete cosine transforms (DCT) domain and are utilized to design the detection scheme. Both theoretical analysis and experimental results show the effectiveness of our proposed scheme on identifying the resampled JPEG images as well as the JPEG images undergone resampling and then JPEG recompression. Moreover, the proposed approach can be used to estimate the resampling factors for restoring the whole operation chain. HighlightsOperation chain consists of JPEG compression and Resampling.Detection relies on TBAG and DCTR.Resampling factor can be estimated.

Still image compression method based on deep convolutional neural network

Abstract: The invention discloses a still image compression method based on a deep convolutional neural network. The still image compression method mainly comprises the following steps of: carrying out downsampling on an original image at an encoding end and carrying out encoding and decoding by utilizing a JPEG2000 standard; carrying out an inhibition compression effect on a decoded image by utilizing the deep convolutional neural network; reconstructing an inhibition compression effect image by adopting a super-resolution method; carrying out subtraction on the original image and a decoded high-resolution image to obtain a residue image and carrying out targeted encoding; forming bit stream by an encoded low-resolution image, the residual image and auxiliary information and transmitting the bit stream; carrying out decoding by a decoding end to obtain a decoded low-resolution image, residual image and auxiliary information; and processing the decoded low-resolution image to obtain a decoded high-resolution image, carrying out superposition on the decoded high-resolution image and the decoded residual image to obtain a finally decoded high-resolution image. The still image compression method disclosed by the invention has higher rate distortion performance than the JPEG2000 standard.

Transfer of Multimedia Data via LoRa

Abstract: This article provides the results of the multimedia data transmission parameters research by LoRa using, in particular the results of images and voice transmission using the fragment of the model network in the Internet of Things Laboratory SPbSUT. During the series of experiments there was noticed the LoRa radio modules performance variation of different parameters (Bandwidth, Spreading Factor and Coding Rate), which have affected on the time and quality of image transmission. For image compression were used JPEG and JPEG 2000 methods, which have allowed to achieve an acceptable compression and image reconstruction while transmitting in the low-speed network. In the course of the experiment, the images were transferred from a camera mounted on a quadrocopter at distance of several kilometers. We considered such parameters as the time of data transfer, packet loss, estimation of the images quality obtained on the basis of subjective and objective methods. For voice compression, the A-law method was used, which allowed to compress the frame size by 4 times. Experiments of real-time speech transmission were conducted in different languages and evaluated by the experts. During the results analyzing there were defined the lower subjective score for Arabic, and the higher scores for English and Vietnamese. In conclusion, this article provides the results of the quantitative and qualitative dimensions evaluation and presents directions for the further research.

A hybrid image compression algorithm based on JPEG and Fuzzy transform

Abstract: We propose a new hybrid image compression algorithm which combines the F-transform and the JPEG. At first, we apply the direct F-transform and then, the JPEG compression. Conversly, the JPEG decompression is followed by the inverse F-transform to obtain the decompressed image. This scheme brings three benefits: (i) the direct F-transform filters out high frequencies so that the JPEG can reach a higher compression ratio; (ii) the JPEG color quantization can be omitted in order to achieve greater decompressed image quality; (iii) the JPEG-decompressed image is processed by by the inverse F-transform w.r.t. the adjoint partition almost lossless. The paper justifies the proposed hybrid algorithm by benchmarks which show that the hybrid algorithm achieves significantly higher decompressed image quality than the JPEG.

Diffusion-Based Inpainting for Coding Remote-Sensing Data

Abstract: Inpainting techniques based on partial differential equations (PDEs), such as diffusion processes, are gaining growing importance as a novel family of image compression methods. Nevertheless, the application of inpainting in the field of hyperspectral imagery has been mainly focused on filling in missing information or dead pixels due to sensor failures. In this letter, we propose a novel PDE-based inpainting algorithm to compress hyperspectral images. The method inpaints separately the known data in the spatial and spectral dimensions. Then, it applies a prediction model to the final inpainting solution to obtain a representation much closer to the original image. Experimental results over a set of hyperspectral images indicate that the proposed algorithm can perform better than a recent proposed extension to prediction-based standard CCSDS-123.0 at low bit rate, better than JPEG 2000 Part 2 with the DWT 9/7 as a spectral transform at all bit rates, and competitive to JPEG 2000 with principal component analysis, the optimal spectral decorrelation transform for Gaussian sources.

A novel high-frequency encoding algorithm for image compression

Abstract: In this paper, a new method for image compression is proposed whose quality is demonstrated through accurate 3D reconstruction from 2D images. The method is based on the discrete cosine transform (DCT) together with a high-frequency minimization encoding algorithm at compression stage and a new concurrent binary search algorithm at decompression stage. The proposed compression method consists of five main steps: (1) divide the image into blocks and apply DCT to each block; (2) apply a high-frequency minimization method to the AC-coefficients reducing each block by 2/3 resulting in a minimized array; (3) build a look up table of probability data to enable the recovery of the original high frequencies at decompression stage; (4) apply a delta or differential operator to the list of DC-components; and (5) apply arithmetic encoding to the outputs of steps (2) and (4). At decompression stage, the look up table and the concurrent binary search algorithm are used to reconstruct all high-frequency AC-coefficients while the DC-components are decoded by reversing the arithmetic coding. Finally, the inverse DCT recovers the original image. We tested the technique by compressing and decompressing 2D images including images with structured light patterns for 3D reconstruction. The technique is compared with JPEG and JPEG2000 through 2D and 3D RMSE. Results demonstrate that the proposed compression method is perceptually superior to JPEG with equivalent quality to JPEG2000. Concerning 3D surface reconstruction from images, it is demonstrated that the proposed method is superior to both JPEG and JPEG2000.

Image Quality Evaluation in Lossy Compressed Images

Abstract: This research focuses on the quantification of image quality in lossy compressed images, exploring the impact of digital artefacts and scene characteristics upon image quality evaluation. A subjective paired comparison test was implemented to assess perceived quality of JPEG 2000 against baseline JPEG over a range of different scene types. Interval scales were generated for both algorithms, which indicated a subjective preference for JPEG 2000, particularly at low bit rates, and these were confirmed by an objective distortion measure. The subjective results did not follow this trend for some scenes however, and both algorithms were found to be scene dependent as a result of the artefacts produced at high compression rates. The scene dependencies were explored from the interval scale results, which allowed scenes to be grouped according to their susceptibilities to each of the algorithms. Groupings were correlated with scene measures applied in a linked study. A pilot study was undertaken to explore perceptibility thresholds of JPEG 2000 of the same set of images. This work was developed with a further experiment to investigate the thresholds of perceptibility and acceptability of higher resolution JPEG 2000 compressed images. A set of images was captured using a professional level full-frame Digital Single Lens Reflex camera, using a raw workflow and carefully controlled image-processing pipeline. The scenes were quantified using a set of simple scene metrics to classify them according to whether they were average, higher than, or lower than average, for a number of scene properties known to affect image compression and perceived image quality; these were used to make a final selection of test images. Image fidelity was investigated using the method of constant stimuli to quantify perceptibility thresholds and just noticeable differences (JNDs) of perceptibility. Thresholds and JNDs of acceptability were also quantified to explore suprathreshold quality evaluation. The relationships between the two thresholds were examined and correlated with the results from the scene measures, to identify more or less susceptible scenes. It was found that the level and differences between the two thresholds was an indicator of scene dependency and could be predicted by certain types of scene characteristics. A third study implemented the soft copy quality ruler as an alternative psychophysical method, by matching the quality of compressed images to a set of images varying in a single attribute, separated by known JND increments of quality. The imaging chain and image processing workflow were evaluated using objective measures of tone reproduction and spatial frequency response. An alternative approach to the creation of ruler images was implemented and tested, and the resulting quality rulers were used to evaluate a subset of the images from the previous study. The quality ruler was found to be successful in identifying scene susceptibilities and observer sensitivity. The fourth investigation explored the implementation of four different image quality metrics. These were the Modular Image Difference Metric, the Structural Similarity Metric, The Multi-scale Structural Similarity Metric and the Weighted Structural Similarity Metric. The metrics were tested against the subjective results and all were found to have linear correlation in terms of predictability of image quality.

Double expanding robust image watermarking based on Spread Spectrum technique and BCH coding

Abstract: This paper presents a new algorithm for blind image watermarking which has a high robustness against common image processing attacks such as noise addition (Gaussian noise, Salt & Pepper noise, Speckle noise and etc.), JPEG and JPEG2000 compressions, Histogram Equalization, Average and Gaussian filters, Scaling and Cropping. According to this fact that a watermark with about 70 bits is enough for copyright protection, consequently in this paper a small watermark (64 bits) have been double expanded into multi larger meaningful bits with applying BCH error correction code and Spread Spectrum technique in order to reduce errors in extraction phase. Approximation subband of two levels DWT transform is divided into non-overlapping blocks and high frequency coefficients of DCT transform of each block is used for embedding the watermark. Embedding technique, which is used in this paper, is Spread Spectrum. Correlation between some coefficients of each embedded block and two predefined groups of random bits is used for watermark extraction, so this method is blind and does not need to the original image or additional information in extraction phase. Another idea, which is used in this paper, is calculating different gain factors for each block of approximation subband according to the texture of each block. Consequently this method allocates smaller gain factors to smooth blocks and larger gain factors to texture and rough blocks. So, manipulating in image will be more robust and imperceptible.

Classification accuracies of malaria infected cells using deep convolutional neural networks based on decompressed images

Abstract: In many biomedical applications, images are stored and transmitted in the form of compressed images. However, typical pattern classifiers are trained using original images. There has been little prior study on how lossily decompressed images would impact the classification performance. In a case study of automatic classification of malaria infected cells, we used decompressed cell images as the inputs to deep convolutional neural networks. We evaluated how various lossy image compression methods and varying compression ratios would impact the classification accuracies. Specifically, we compared four compression methods: lossy compression via bitplane reduction, JPEG and JPEG 2000, and sparse autoencoders. Decompressed images were fed into LeNet-5 for training and testing. Simulation results showed that for similar compression ratios, the bitplane reduction method had the lowest classification accuracy, while JPEG and JPEG 2000 methods could maintain good accuracies. In particular, JPEG 2000 decompressed images could achieve about 95% accuracy even after 30 to 1 compression. We also provide classification results based on the widely used MNIST dataset, where handwritten digits were found to be much easier to classify using decompressed images, with about 90% accuracy still achievable using only one single bitplane. As a lossy compression method, Autoencoder was also applied to the MNIST dataset, achieving about 85% accuracy with a compression ratio much higher than the other three lossy image compression methods. Autoencoders were also found to provide more scalable compression ratios, while capable of maintaining good classification accuracies.