Search results for: PSNR MSE

Authors: Khader S. Al-Aidmat, Venus W. Samawi

Abstract:

Blurred vision in the misty atmosphere is essential problem which needs to be resolved. To solve this problem, we developed a technique to restore foggy degraded image from its original version using Back-propagation neural network (BP-NN). The suggested technique is based on mapping between foggy scene and its corresponding original scene. Seven different approaches are suggested based on type of features used in image restoration. Features are extracted from spatial and spatial-frequency domain (using DCT). Each of these approaches comes with its own BP-NN architecture depending on type and number of used features. The weight matrix resulted from training each BP-NN represents a fog filter. The performance of these filters are evaluated empirically (using PSNR), and perceptually. By comparing the performance of these filters, the effective features that suits BP-NN technique for restoring foggy images is recognized. This system proved its effectiveness and success in restoring moderate foggy images.

Keywords: artificial neural network, discrete cosine transform, feed forward neural network, foggy image restoration

Procedia PDF Downloads 380

Authors: Hamza A. Al-Sewadi, Akram N. A. Aldakari

Abstract:

Copyright protection and ownership proof of digital multimedia are achieved nowadays by digital watermarking techniques. A text watermarking algorithm for protecting the property rights and ownership judgment of color images is proposed in this paper. Embedding is achieved by inserting texts elements randomly into the color image as noise. The YIQ image processing model is found to be faster than other image processing methods, and hence, it is adopted for the embedding process. An optional choice of encrypting the text watermark before embedding is also suggested (in case required by some applications), where, the text can is encrypted using any enciphering technique adding more difficulty to hackers. Experiments resulted in embedding speed improvement of more than double the speed of other considered systems (such as least significant bit method, and separate color code methods), and a fairly acceptable level of peak signal to noise ratio (PSNR) with low mean square error values for watermarking purposes.

Keywords: steganography, watermarking, time complexity measurements, private keys

Procedia PDF Downloads 142

Authors: Tanish Mittal, Preyansh Agrawal, Esha Pahwa, Aarya Makwana

Abstract:

Multi-Scale and U-shaped Networks are widely used in various image restoration problems, including deblurring. Keeping in mind the wide range of applications, we present a comparison of these architectures and their effects on image deblurring. We also introduce a new block called as NFResblock. It consists of a Fast Fourier Transformation layer and a series of modified Non-Linear Activation Free Blocks. Based on these architectures and additions, we introduce NFResnet and NFResnet+, which are modified multi-scale and U-Net architectures, respectively. We also use three differ-ent loss functions to train these architectures: Charbonnier Loss, Edge Loss, and Frequency Reconstruction Loss. Extensive experiments on the Deep Video Deblurring dataset, along with ablation studies for each component, have been presented in this paper. The proposed architectures achieve a considerable increase in Peak Signal to Noise (PSNR) ratio and Structural Similarity Index (SSIM) value.

Keywords: multi-scale, Unet, deblurring, FFT, resblock, NAF-block, nfresnet, charbonnier, edge, frequency reconstruction

Procedia PDF Downloads 135

Authors: Volkan Kaya, Ersin Elbasi

Abstract:

Watermarking and stenography are getting importance recently because of copyright protection and authentication. In watermarking we embed stamp, logo, noise or image to multimedia elements such as image, video, audio, animation and text. There are several works have been done in watermarking for different purposes. In this research work, we used watermarking techniques to embed patient information into the medical magnetic resonance (MR) images. There are two methods have been used; frequency domain (Digital Wavelet Transform-DWT, Digital Cosine Transform-DCT, and Digital Fourier Transform-DFT) and spatial domain (Least Significant Bits-LSB) domain. Experimental results show that embedding in frequency domains resist against one type of attacks, and embedding in spatial domain is resist against another group of attacks. Peak Signal Noise Ratio (PSNR) and Similarity Ratio (SR) values are two measurement values for testing. These two values give very promising result for information hiding in medical MR images.

Keywords: watermarking, medical image, frequency domain, least significant bits, security

Procedia PDF Downloads 286

Authors: Amhimmid Mohammed Saffour, Mustafa Mohamed Abdullah

Abstract:

The use of computers for handling image data in the healthcare is growing. However, the amount of data produced by modern image generating techniques is vast. This data might be a problem from a storage point of view or when the data is sent over a network. This paper using wavelet transform technique for medical images compression. MATLAB program, are designed to evaluate medical images storage and transmission time problem at Sebha Medical Center Libya. In this paper, three different Computed Tomography images which are abdomen, brain and chest have been selected and compressed using wavelet transform. Objective evaluation has been performed to measure the quality of the compressed images. For this evaluation, the results show that the Peak Signal to Noise Ratio (PSNR) which indicates the quality of the compressed image is ranging from (25.89db to 34.35db for abdomen images, 23.26db to 33.3db for brain images and 25.5db to 36.11db for chest images. These values shows that the compression ratio is nearly to 30:1 is acceptable.

Keywords: medical image, Matlab, image compression, wavelet's, objective evaluation

Procedia PDF Downloads 284

Authors: Khalid A. Al-Afandy, El-Sayyed El-Rabaie, Osama Salah, Ahmed El-Mhalaway

Abstract:

This paper presents a high secure data hiding technique using image cropping and Least Significant Bit (LSB) steganography. The predefined certain secret coordinate crops will be extracted from the cover image. The secret text message will be divided into sections. These sections quantity is equal the image crops quantity. Each section from the secret text message will embed into an image crop with a secret sequence using LSB technique. The embedding is done using the cover image color channels. Stego image is given by reassembling the image and the stego crops. The results of the technique will be compared to the other state of art techniques. Evaluation is based on visualization to detect any degradation of stego image, the difficulty of extracting the embedded data by any unauthorized viewer, Peak Signal-to-Noise Ratio of stego image (PSNR), and the embedding algorithm CPU time. Experimental results ensure that the proposed technique is more secure compared with the other traditional techniques.

Keywords: steganography, stego, LSB, crop

Procedia PDF Downloads 267

Authors: M. V. Chidananda Murthy, M. Z. Kurian, H. S. Guruprasad

Abstract:

Image restoration in the process of communication is one of the emerging fields in the image processing. The motion analysis processing is the simplest case to detect motion in an image. Applications of motion analysis widely spread in many areas such as surveillance, remote sensing, film industry, navigation of autonomous vehicles, etc. The scene may contain multiple moving objects, by using motion analysis techniques the blur caused by the movement of the objects can be enhanced by filling-in occluded regions and reconstruction of transparent objects, and it also removes the motion blurring. This paper presents the design and comparison of various motion detection and enhancement filters. Median filter, Linear image deconvolution, Inverse filter, Pseudoinverse filter, Wiener filter, Lucy Richardson filter and Blind deconvolution filters are used to remove the blur. In this work, we have considered different types and different amount of blur for the analysis. Mean Square Error (MSE) and Peak Signal to Noise Ration (PSNR) are used to evaluate the performance of the filters. The designed system has been implemented in Matlab software and tested for synthetic and real-time images.

Keywords: image enhancement, motion analysis, motion detection, motion estimation

Procedia PDF Downloads 284

Authors: Toni Maristela C. Estabillo, Michaela V. Matienzo, Mikaela L. Sabangan, Rosette M. Tienzo, Justine L. Bahinting

Abstract:

This study is about blind watermarking on images with different categories and properties using two algorithms namely, Discrete Wavelet Transform and Patchwork Algorithm. A program is created to perform watermark embedding, extraction and evaluation. The evaluation is based on three watermarking criteria namely: image quality degradation, perceptual transparency and security. Image quality is measured by comparing the original properties with the processed one. Perceptual transparency is measured by a visual inspection on a survey. Security is measured by implementing geometrical and non-geometrical attacks through a pass or fail testing. Values used to measure the following criteria are mostly based on Mean Squared Error (MSE) and Peak Signal to Noise Ratio (PSNR). The results are based on statistical methods used to interpret and collect data such as averaging, z Test and survey. The study concluded that the combined DWT and Patchwork algorithms were less efficient and less capable of watermarking than DWT algorithm only.

Keywords: blind watermarking, discrete wavelet transform algorithm, patchwork algorithm, digital watermark

Procedia PDF Downloads 267

Authors: Amar Partap Singh Pharwaha, Balkrishan Jindal

Abstract:

In this paper, a new method is developed for hiding image in a digital image with multilayer security. In the proposed method, the secret image is encrypted in the first instance using a flexible matrix based symmetric key to add first layer of security. Then another layer of security is added to the secret data by encrypting the ciphered data using Pythagorean Theorem method. The ciphered data bits (4 bits) produced after double encryption are then embedded within digital image in the spatial domain using Least Significant Bits (LSBs) substitution. To improve the image quality of the stego-image, an improved form of pixel adjustment process is proposed. To evaluate the effectiveness of the proposed method, image quality metrics including Peak Signal-to-Noise Ratio (PSNR), Mean Square Error (MSE), entropy, correlation, mean value and Universal Image Quality Index (UIQI) are measured. It has been found experimentally that the proposed method provides higher security as well as robustness. In fact, the results of this study are quite promising.

Keywords: Pythagorean theorem, pixel adjustment, ciphered data, image hiding, least significant bit, flexible matrix

Procedia PDF Downloads 335

Authors: Harsh Vikram Singh, Ankur Rai, Anand Mohan

Abstract:

In this paper, we propose secure and robust data hiding algorithm based on DCT by Arnold transform and chaotic sequence. The watermark image is scrambled by Arnold cat map to increases its security and then the chaotic map is used for watermark signal spread in middle band of DCT coefficients of the cover image The chaotic map can be used as pseudo-random generator for digital data hiding, to increase security and robustness .Performance evaluation for robustness and imperceptibility of proposed algorithm has been made using bit error rate (BER), normalized correlation (NC), and peak signal to noise ratio (PSNR) value for different watermark and cover images such as Lena, Girl, Tank images and gain factor .We use a binary logo image and text image as watermark. The experimental results demonstrate that the proposed algorithm achieves higher security and robustness against JPEG compression as well as other attacks such as addition of noise, low pass filtering and cropping attacks compared to other existing algorithm using DCT coefficients. Moreover, to recover watermarks in proposed algorithm, there is no need to original cover image.

Keywords: data hiding, watermarking, DCT, chaotic sequence, arnold transforms

Procedia PDF Downloads 512

Authors: Chafik Barnoussi, Mourad Talbi, Adnane Cherif

Abstract:

In this paper we propose a new speech compression system based on the application of the Bionic Wavelet Transform (BWT) combined with the psychoacoustic model. This compression system is a modified version of the compression system using a MDCT (Modified Discrete Cosine Transform) filter banks of 32 filters each and the psychoacoustic model. This modification consists in replacing the banks of the MDCT filter banks by the bionic wavelet coefficients which are obtained from the application of the BWT to the speech signal to be compressed. These two methods are evaluated and compared with each other by computing bits before and bits after compression. They are tested on different speech signals and the obtained simulation results show that the proposed technique outperforms the second technique and this in term of compressed file size. In term of SNR, PSNR and NRMSE, the outputs speech signals of the proposed compression system are with acceptable quality. In term of PESQ and speech signal intelligibility, the proposed speech compression technique permits to obtain reconstructed speech signals with good quality.

Keywords: speech compression, bionic wavelet transform, filterbanks, psychoacoustic model

Procedia PDF Downloads 381

Authors: Biji Jacob

Abstract:

Vector quantization algorithm which uses minimum distance calculation for codebook generation, a time consuming calculation performed on each pixel values leads to computation complexity. The codebook is updated by comparing the distance of each vector to their centroid vector and measure for their closeness. In this paper vector quantization is modified based on vector difference algorithm for image enhancement purpose. In the proposed scheme, vector differences between the vectors are considered as the new generation vectors or new codebook vectors. The codebook is updated by comparing the new generation vector with a threshold value having minimum error with the parent vector. The minimum error decides the fitness of each newly generated vector. Thus the codebook is generated in an adaptive manner and the fitness value is determined for the suppression of the degraded portion of the image and thereby leads to the enhancement of the image through the adaptive searching capability of the vector quantization through vector difference algorithm. Experimental results shows that the vector difference scheme efficiently modifies the vector quantization algorithm for enhancing the image with peak signal to noise ratio (PSNR), mean square error (MSE), Euclidean distance (E_dist) as the performance parameters.

Keywords: codebook, image enhancement, vector difference, vector quantization

Procedia PDF Downloads 264

Authors: R. Rama Kishore, Sunesh

Abstract:

This paper presents an optimized, robust, and secured watermarking technique. The methodology used in this work is the combination of entropy and chaotic grid map. The proposed methodology incorporates Discrete Cosine Transform (DCT) on the host image. To improve the imperceptibility of the method, the host image DCT blocks, where the watermark is to be embedded, are further optimized by considering the entropy of the blocks. Chaotic grid is used as a key to reorder the DCT blocks so that it will further increase security while selecting the watermark embedding locations and its sequence. Without a key, one cannot reveal the exact watermark from the watermarked image. The proposed method is implemented on four different images. It is concluded that the proposed method is giving better results in terms of imperceptibility measured through PSNR and found to be above 50. In order to prove the effectiveness of the method, the performance analysis is done after implementing different attacks on the watermarked images. It is found that the methodology is very strong against JPEG compression attack even with the quality parameter up to 15. The experimental results are confirming that the combination of entropy and chaotic grid map method is strong and secured to different image processing attacks.

Keywords: digital watermarking, discreate cosine transform, chaotic grid map, entropy

Procedia PDF Downloads 252

Authors: Shahriar Farzam, Maryam Rastgarpour

Abstract:

Image denoising plays extremely important role in digital image processing. Enhancement of clinical image research based on Curvelet has been developed rapidly in recent years. In this paper, we present a method for image contrast enhancement for cone beam CT (CBCT) images based on fast discrete curvelet transforms (FDCT) that work through Unequally Spaced Fast Fourier Transform (USFFT). These transforms return a table of Curvelet transform coefficients indexed by a scale parameter, an orientation and a spatial location. Accordingly, the coefficients obtained from FDCT-USFFT can be modified in order to enhance contrast in an image. Our proposed method first uses a two-dimensional mathematical transform, namely the FDCT through unequal-space fast Fourier transform on input image and then applies thresholding on coefficients of Curvelet to enhance the CBCT images. Consequently, applying unequal-space fast Fourier Transform leads to an accurate reconstruction of the image with high resolution. The experimental results indicate the performance of the proposed method is superior to the existing ones in terms of Peak Signal to Noise Ratio (PSNR) and Effective Measure of Enhancement (EME).

Keywords: curvelet transform, CBCT, image enhancement, image denoising

Procedia PDF Downloads 299

Authors: Harshan Baskar, Anirudh S. Chakravarthy, Prateek Garg, Divyam Goel, Abhijith S. Raj, Kshitij Kumar, Lakshya, Ravichandra Parvatham, V. Sushant, Bijay Kumar Rout

Abstract:

In this paper, we introduce a new computer vision task called nighttime dehaze-enhancement. This task aims to jointly perform dehazing and lightness enhancement. Our task fundamentally differs from nighttime dehazing – our goal is to jointly dehaze and enhance scenes, while nighttime dehazing aims to dehaze scenes under a nighttime setting. In order to facilitate further research on this task, we release a new benchmark dataset called Reside-β Night dataset, consisting of 4122 nighttime hazed images from 2061 scenes and 2061 ground truth images. Moreover, we also propose a new network called NDENet (Nighttime Dehaze-Enhancement Network), which jointly performs dehazing and low-light enhancement in an end-to-end manner. We evaluate our method on the proposed benchmark and achieve SSIM of 0.8962 and PSNR of 26.25. We also compare our network with other baseline networks on our benchmark to demonstrate the effectiveness of our approach. We believe that nighttime dehaze-enhancement is an essential task, particularly for autonomous navigation applications, and we hope that our work will open up new frontiers in research. Our dataset and code will be made publicly available upon acceptance of our paper.

Keywords: dehazing, image enhancement, nighttime, computer vision

Procedia PDF Downloads 157

Authors: Olaiya Folorunsho, Comfort Y. Daramola, Joel N. Ugwu, Lawrence B. Adewole, Olufisayo S. Ekundayo

Abstract:

In today world, there is a tremendous rise in the usage of internet due to the fact that almost all the communication and information sharing is done over the web. Conversely, there is a continuous growth of unauthorized access to confidential data. This has posed a challenge to information security expertise whose major goal is to curtail the menace. One of the approaches to secure the safety delivery of data/information to the rightful destination without any modification is steganography. Steganography is the art of hiding information inside an embedded information. This research paper aimed at designing a secured algorithm with the use of image steganographic technique that makes use of Least Significant Bit (LSB) algorithm for embedding the data into the bit map image (bmp) in order to enhance security and reliability. In the LSB approach, the basic idea is to replace the LSB of the pixels of the cover image with the Bits of the messages to be hidden without destroying the property of the cover image significantly. The system was implemented using C# programming language of Microsoft.NET framework. The performance evaluation of the proposed system was experimented by conducting a benchmarking test for analyzing the parameters like Mean Squared Error (MSE) and Peak Signal to Noise Ratio (PSNR). The result showed that image steganography performed considerably in securing data hiding and information transmission over the networks.

Keywords: steganography, image steganography, least significant bits, bit map image

Procedia PDF Downloads 266

Authors: Cherifi Mehdi, Lahdir Mourad, Ameur Soltane

Abstract:

Image compression is used to reduce the number of bits required to represent an image. The Meteosat Second Generation satellite (MSG) allows the acquisition of 12 image files every 15 minutes. Which results a large databases sizes. The transform selected in the images compression should contribute to reduce the data representing the images. The Radon transform retrieves the Radon points that represent the sum of the pixels in a given angle for each direction. Linear predictive coding (LPC) with filtering provides a good decorrelation of Radon points using a Predictor constitute by the Symmetric Nearest Neighbor filter (SNN) coefficients, which result losses during decompression. Finally, Run Length Coding (RLC) gives us a high and fixed compression ratio regardless of the input image. In this paper, a novel image compression method based on the Radon transform and linear predictive coding (LPC) for MSG images is proposed. MSG image compression based on the Radon transform and the LPC provides a good compromise between compression and quality of reconstruction. A comparison of our method with other whose two based on DCT and one on DWT bi-orthogonal filtering is evaluated to show the power of the Radon transform in its resistibility against the quantization noise and to evaluate the performance of our method. Evaluation criteria like PSNR and the compression ratio allows showing the efficiency of our method of compression.

Keywords: image compression, radon transform, linear predictive coding (LPC), run lengthcoding (RLC), meteosat second generation (MSG)

Procedia PDF Downloads 418

Authors: Xiaolin Zeng, Lei Cheng, Zhirong Li, Xueping Liu

Abstract:

Aiming at the problem of low accuracy and susceptibility to cultural relic diseases in the generation of high-resolution archaeology drawings by current image generation algorithms, an archaeology drawings generation algorithm based on a conditional generative adversarial network is proposed. An attention mechanism is added into the high-resolution image generation network as the backbone network, which enhances the line feature extraction capability and improves the accuracy of line drawing generation. A dual-branch parallel architecture consisting of two backbone networks is implemented, where the semantic translation branch extracts semantic features from orthophotographs of cultural relics, and the gradient screening branch extracts effective gradient features. Finally, the fusion fine-tuning module combines these two types of features to achieve the generation of high-quality and high-resolution archaeology drawings. Experimental results on the self-constructed archaeology drawings dataset of grotto temple statues show that the proposed algorithm outperforms current mainstream image generation algorithms in terms of pixel accuracy (PA), structural similarity (SSIM), and peak signal-to-noise ratio (PSNR) and can be used to assist in drawing archaeology drawings.

Keywords: archaeology drawings, digital heritage, image generation, deep learning

Procedia PDF Downloads 56

Authors: Deng Zengming, Wang Mingjiang

Abstract:

As 3D video is explored as a hot research topic in the last few decades, free-viewpoint TV (FTV) is no doubt a promising field for its better visual experience and incomparable interactivity. View synthesis is obviously a crucial technology for FTV; it enables to render images in unlimited numbers of virtual viewpoints with the information from limited numbers of reference view. In this paper, a novel hybrid synthesis framework is proposed and blending priority is explored. In contrast to the commonly used View Synthesis Reference Software (VSRS), the presented synthesis process is driven in consideration of the temporal correlation of image sequences. The temporal correlations will be exploited to produce fine synthesis results even near the foreground boundaries. As for the blending priority, this scheme proposed that one of the two reference views is selected to be the main reference view based on the distance between the reference views and virtual view, another view is chosen as the auxiliary viewpoint, just assist to fill the hole pixel with the help of background information. Significant improvement of the proposed approach over the state-of –the-art pixel-based virtual view synthesis method is presented, the results of the experiments show that subjective gains can be observed, and objective PSNR average gains range from 0.5 to 1.3 dB, while SSIM average gains range from 0.01 to 0.05.

Keywords: fusion method, Gaussian mixture model, hybrid framework, view synthesis

Procedia PDF Downloads 248

Authors: RAM PAL SINGH, VIKASH CHAUDHARY, MONIKA VERMA

Abstract:

In this paper, we proposed the implementation of optimization based Extreme Learning Machine (ELM) for watermarking of B-channel of color image in discrete wavelet transform (DWT) domain. ELM, a regularization algorithm, works based on generalized single-hidden-layer feed-forward neural networks (SLFNs). However, hidden layer parameters, generally called feature mapping in context of ELM need not to be tuned every time. This paper shows the embedding and extraction processes of watermark with the help of ELM and results are compared with already used machine learning models for watermarking.Here, a cover image is divide into suitable numbers of non-overlapping blocks of required size and DWT is applied to each block to be transformed in low frequency sub-band domain. Basically, ELM gives a unified leaning platform with a feature mapping, that is, mapping between hidden layer and output layer of SLFNs, is tried for watermark embedding and extraction purpose in a cover image. Although ELM has widespread application right from binary classification, multiclass classification to regression and function estimation etc. Unlike SVM based algorithm which achieve suboptimal solution with high computational complexity, ELM can provide better generalization performance results with very small complexity. Efficacy of optimization method based ELM algorithm is measured by using quantitative and qualitative parameters on a watermarked image even though image is subjected to different types of geometrical and conventional attacks.

Keywords: BER, DWT, extreme leaning machine (ELM), PSNR

Procedia PDF Downloads 310

Authors: Chia Jui Hsieh, Jyh Cheng Chen, Chih Wei Kuo, Ruei Teng Wang, Woei Chyn Chu

Abstract:

Compressed sensing (CS) based computed tomographic (CT) reconstruction algorithm utilizes total variation (TV) to transform CT image into sparse domain and minimizes L1-norm of sparse image for reconstruction. Different from the traditional CS based reconstruction which only calculates x-coordinate and y-coordinate TV to transform CT images into sparse domain, we propose a multi-directional TV to transform tomographic image into sparse domain for low-dose reconstruction. Our method considers all possible directions of TV calculations around a pixel, so the sparse transform for CS based reconstruction is more accurate. In 2D CT reconstruction, we use eight-directional TV to transform CT image into sparse domain. Furthermore, we also use 26-directional TV for 3D reconstruction. This multi-directional sparse transform method makes CS based reconstruction algorithm more powerful to reduce noise and increase image quality. To validate and evaluate the performance of this multi-directional sparse transform method, we use both Shepp-Logan phantom and a head phantom as the targets for reconstruction with the corresponding simulated sparse projection data (angular sampling interval is 5 deg and 6 deg, respectively). From the results, the multi-directional TV method can reconstruct images with relatively less artifacts compared with traditional CS based reconstruction algorithm which only calculates x-coordinate and y-coordinate TV. We also choose RMSE, PSNR, UQI to be the parameters for quantitative analysis. From the results of quantitative analysis, no matter which parameter is calculated, the multi-directional TV method, which we proposed, is better.

Keywords: compressed sensing (CS), low-dose CT reconstruction, total variation (TV), multi-directional gradient operator

Procedia PDF Downloads 255

Authors: Linyu Wang, Liguo Qiao, Jianhong Xiang, Hao Xu

Abstract:

In semantic communication, the existing joint Source Channel coding (JSCC) wireless communication system without pilot has unstable transmission performance and can not effectively capture the global information and location information of images. In this paper, a pilot-free image transmission system of joint source channel based on multi-level semantic information (Multi-level JSCC) is proposed. The transmitter of the system is composed of two networks. The feature extraction network is used to extract the high-level semantic features of the image, compress the information transmitted by the image, and improve the bandwidth utilization. Feature retention network is used to preserve low-level semantic features and image details to improve communication quality. The receiver also is composed of two networks. The received high-level semantic features are fused with the low-level semantic features after feature enhancement network in the same dimension, and then the image dimension is restored through feature recovery network, and the image location information is effectively used for image reconstruction. This paper verifies that the proposed multi-level JSCC algorithm can effectively transmit and recover image information in both AWGN channel and Rayleigh fading channel, and the peak signal-to-noise ratio (PSNR) is improved by 1~2dB compared with other algorithms under the same simulation conditions.

Keywords: deep learning, JSCC, pilot-free picture transmission, multilevel semantic information, robustness

Procedia PDF Downloads 120

Authors: Navdeep Goel, Salvador Gabarda

Abstract:

Digital images are widely used in computer applications. To store or transmit the uncompressed images requires considerable storage capacity and transmission bandwidth. Image compression is a means to perform transmission or storage of visual data in the most economical way. This paper explains about how images can be encoded to be transmitted in a multiplexing time-frequency domain channel. Multiplexing involves packing signals together whose representations are compact in the working domain. In order to optimize transmission resources each 4x4 pixel block of the image is transformed by a suitable polynomial approximation, into a minimal number of coefficients. Less than 4*4 coefficients in one block spares a significant amount of transmitted information, but some information is lost. Different approximations for image transformation have been evaluated as polynomial representation (Vandermonde matrix), least squares + gradient descent, 1-D Chebyshev polynomials, 2-D Chebyshev polynomials or singular value decomposition (SVD). Results have been compared in terms of nominal compression rate (NCR), compression ratio (CR) and peak signal-to-noise ratio (PSNR) in order to minimize the error function defined as the difference between the original pixel gray levels and the approximated polynomial output. Polynomial coefficients have been later encoded and handled for generating chirps in a target rate of about two chirps per 4*4 pixel block and then submitted to a transmission multiplexing operation in the time-frequency domain.

Keywords: chirp signals, image multiplexing, image transformation, linear canonical transform, polynomial approximation

Procedia PDF Downloads 410

Authors: Hussain Sajid, Jung-Hun Shin, Kum-Won Cho

Abstract:

Image super-resolution is the most common computer vision problem with many important applications. Generative adversarial networks (GANs) have promoted remarkable advances in single-image super-resolution (SR) by recovering photo-realistic images. However, high memory requirements of GAN-based SR (mainly generators) lead to performance degradation and increased energy consumption, making it difficult to implement it onto resource-constricted devices. To relieve such a problem, In this paper, we introduce an optimized and highly efficient architecture for SR-GAN (generator) model by utilizing model compression techniques such as Knowledge Distillation and pruning, which work together to reduce the storage requirement of the model also increase in their performance. Our method begins with distilling the knowledge from a large pre-trained model to a lightweight model using different loss functions. Then, iterative weight pruning is applied to the distilled model to remove less significant weights based on their magnitude, resulting in a sparser network. Knowledge Distillation reduces the model size by 40%; pruning then reduces it further by 18%. To accelerate the learning process, we employ the Horovod framework for distributed training on a cluster of 2 nodes, each with 8 GPUs, resulting in improved training performance and faster convergence. Experimental results on various benchmarks demonstrate that the proposed compressed model significantly outperforms state-of-the-art methods in terms of peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and image quality for x4 super-resolution tasks.

Keywords: single-image super-resolution, generative adversarial networks, knowledge distillation, pruning

Procedia PDF Downloads 95

Authors: Abdelhalim. N. Mohammed, Mohammed. Y. Esmail

Abstract:

In the course of recent decades, medical imaging has been dominated by the use of costly film media for review and archival of medical investigation, however due to developments in networks technologies and common acceptance of a standard digital imaging and communication in medicine (DICOM) another approach in light of World Wide Web was produced. Web technologies successfully used in telemedicine applications, the combination of web technologies together with DICOM used to design a web-based and open source DICOM viewer. The Web server allowance to inquiry and recovery of images and the images viewed/manipulated inside a Web browser without need for any preinstalling software. The dynamic site page for medical images visualization and processing created by using JavaScript and HTML5 advancements. The XAMPP ‘apache server’ is used to create a local web server for testing and deployment of the dynamic site. The web-based viewer connected to multiples devices through local area network (LAN) to distribute the images inside healthcare facilities. The system offers a few focal points over ordinary picture archiving and communication systems (PACS): easy to introduce, maintain and independently platforms that allow images to display and manipulated efficiently, the system also user-friendly and easy to integrate with an existing system that have already been making use of web technologies. The wavelet-based image compression technique on which 2-D discrete wavelet transform used to decompose the image then wavelet coefficients are transmitted by entropy encoding after threshold to decrease transmission time, stockpiling cost and capacity. The performance of compression was estimated by using images quality metrics such as mean square error ‘MSE’, peak signal to noise ratio ‘PSNR’ and compression ratio ‘CR’ that achieved (83.86%) when ‘coif3’ wavelet filter is used.

Keywords: DICOM, discrete wavelet transform, PACS, HIS, LAN

Procedia PDF Downloads 160

Authors: Matheus Madureira Matos, Alexandre Rodrigues Farias

Abstract:

Magnetic Resonance Imaging (MRI) is a vital imaging technique used in both clinical and pre-clinical areas to obtain detailed anatomical and functional information. However, MRI scans can be expensive, time-consuming, and often require the use of anesthetics to keep animals still during the imaging process. Anesthetics are commonly administered to animals undergoing MRI scans to ensure they remain still during the imaging process. However, prolonged or repeated exposure to anesthetics can have adverse effects on animals, including physiological alterations and potential toxicity. Minimizing the duration and frequency of anesthesia is, therefore, crucial for the well-being of research animals. In recent years, various sampling trajectories have been investigated to reduce the number of MRI measurements leading to shorter scanning time and minimizing the duration of animal exposure to the effects of anesthetics. Compressed sensing (CS) and sampling trajectories, such as cartesian, spiral, and radial, have emerged as powerful tools to reduce MRI data while preserving diagnostic quality. This work aims to apply CS and cartesian, spiral, and radial sampling trajectories for the reconstruction of MRI of the abdomen of mice sub-sampled at levels below that defined by the Nyquist theorem. The methodology of this work consists of using a fully sampled reference MRI of a female model C57B1/6 mouse acquired experimentally in a 4.7 Tesla MRI scanner for small animals using Spin Echo pulse sequences. The image is down-sampled by cartesian, radial, and spiral sampling paths and then reconstructed by CS. The quality of the reconstructed images is objectively assessed by three quality assessment techniques RMSE (Root mean square error), PSNR (Peak to Signal Noise Ratio), and SSIM (Structural similarity index measure). The utilization of optimized sampling trajectories and CS technique has demonstrated the potential for a significant reduction of up to 70% of image data acquisition. This result translates into shorter scan times, minimizing the duration and frequency of anesthesia administration and reducing the potential risks associated with it.

Keywords: compressed sensing, magnetic resonance, sampling trajectories, small animals

Procedia PDF Downloads 72

Authors: Zhao Gao, Eran Edirisinghe

Abstract:

The production of an accurate sketch of a suspect based on a verbal description obtained from a witness is an essential task for most criminal investigations. The criminal investigation system employs specifically trained professional artists to manually draw a facial image of the suspect according to the descriptions of an eyewitness for subsequent identification. Within the advancement of Deep Learning, Recurrent Neural Networks (RNN) have shown great promise in Natural Language Processing (NLP) tasks. Additionally, Generative Adversarial Networks (GAN) have also proven to be very effective in image generation. In this study, a trained GAN conditioned on textual features such as keywords automatically encoded from a verbal description of a human face using an RNN is used to generate photo-realistic facial images for criminal investigations. The intention of the proposed system is to map corresponding features into text generated from verbal descriptions. With this, it becomes possible to generate many reasonably accurate alternatives to which the witness can use to hopefully identify a suspect from. This reduces subjectivity in decision making both by the eyewitness and the artist while giving an opportunity for the witness to evaluate and reconsider decisions. Furthermore, the proposed approach benefits law enforcement agencies by reducing the time taken to physically draw each potential sketch, thus increasing response times and mitigating potentially malicious human intervention. With publically available 'CelebFaces Attributes Dataset' (CelebA) and additionally providing verbal description as training data, the proposed architecture is able to effectively produce facial structures from given text. Word Embeddings are learnt by applying the RNN architecture in order to perform semantic parsing, the output of which is fed into the GAN for synthesizing photo-realistic images. Rather than the grid search method, a metaheuristic search based on genetic algorithms is applied to evolve the network with the intent of achieving optimal hyperparameters in a fraction the time of a typical brute force approach. With the exception of the ‘CelebA’ training database, further novel test cases are supplied to the network for evaluation. Witness reports detailing criminals from Interpol or other law enforcement agencies are sampled on the network. Using the descriptions provided, samples are generated and compared with the ground truth images of a criminal in order to calculate the similarities. Two factors are used for performance evaluation: The Structural Similarity Index (SSIM) and the Peak Signal-to-Noise Ratio (PSNR). A high percentile output from this performance matrix should attribute to demonstrating the accuracy, in hope of proving that the proposed approach can be an effective tool for law enforcement agencies. The proposed approach to criminal facial image generation has potential to increase the ratio of criminal cases that can be ultimately resolved using eyewitness information gathering.

Keywords: RNN, GAN, NLP, facial composition, criminal investigation

Procedia PDF Downloads 159

Authors: Hyohun Kim, Dongwha Shin, Yeonseok Kim, Ji-Su Ahn, Kensuke Nakamura, Dongeun Choi, Byung-Woo Hong

Abstract:

Artefacts are commonly encountered in the imaging process of clinical computed tomography (CT) where the artefact refers to any systematic discrepancy between the reconstructed observation and the true attenuation coefficient of the object. It is known that CT images are inherently more prone to artefacts due to its image formation process where a large number of independent detectors are involved, and they are assumed to yield consistent measurements. There are a number of different artefact types including noise, beam hardening, scatter, pseudo-enhancement, motion, helical, ring, and metal artefacts, which cause serious difficulties in reading images. Thus, it is desired to remove nuisance factors from the degraded image leaving the fundamental intrinsic information that can provide better interpretation of the anatomical and pathological characteristics. However, it is considered as a difficult task due to the high dimensionality and variability of data to be recovered, which naturally motivates the use of machine learning techniques. We propose an image restoration algorithm based on the deep neural network framework where the denoising auto-encoders are stacked building multiple layers. The denoising auto-encoder is a variant of a classical auto-encoder that takes an input data and maps it to a hidden representation through a deterministic mapping using a non-linear activation function. The latent representation is then mapped back into a reconstruction the size of which is the same as the size of the input data. The reconstruction error can be measured by the traditional squared error assuming the residual follows a normal distribution. In addition to the designed loss function, an effective regularization scheme using residual-driven dropout determined based on the gradient at each layer. The optimal weights are computed by the classical stochastic gradient descent algorithm combined with the back-propagation algorithm. In our algorithm, we initially decompose an input image into its intrinsic representation and the nuisance factors including artefacts based on the classical Total Variation problem that can be efficiently optimized by the convex optimization algorithm such as primal-dual method. The intrinsic forms of the input images are provided to the deep denosing auto-encoders with their original forms in the training phase. In the testing phase, a given image is first decomposed into the intrinsic form and then provided to the trained network to obtain its reconstruction. We apply our algorithm to the restoration of the corrupted CT images by the artefacts. It is shown that our algorithm improves the readability and enhances the anatomical and pathological properties of the object. The quantitative evaluation is performed in terms of the PSNR, and the qualitative evaluation provides significant improvement in reading images despite degrading artefacts. The experimental results indicate the potential of our algorithm as a prior solution to the image interpretation tasks in a variety of medical imaging applications. This work was supported by the MISP(Ministry of Science and ICT), Korea, under the National Program for Excellence in SW (20170001000011001) supervised by the IITP(Institute for Information and Communications Technology Promotion).

Keywords: auto-encoder neural network, CT image artefact, deep learning, intrinsic image representation, noise reduction, total variation

Procedia PDF Downloads 189