PSNR Results Research Articles

A Novel Medical Image Denoising Method Based on Conditional Generative Adversarial Network.

Medical image quality is highly relative to clinical diagnosis and treatment, leading to a popular research topic of medical image denoising. Image denoising based on deep learning methods has attracted considerable attention owing to its excellent ability of automatic feature extraction. Most existing methods for medical image denoising adapted to certain types of noise have difficulties in handling spatially varying noise; meanwhile, image detail losses and structure changes occurred in the denoised image. Considering image context perception and structure preserving, this paper firstly introduces a medical image denoising method based on conditional generative adversarial network (CGAN) for various unknown noises. In the proposed architecture, noise image with the corresponding gradient image is merged as network conditional information, which enhances the contrast between the original signal and noise according to the structural specificity. A novel generator with residual dense blocks makes full use of the relationship among convolutional layers to explore image context. Furthermore, the reconstruction loss and WGAN loss are combined as the objective loss function to ensure the consistency of denoised image and real image. A series of experiments for medical image denoising are conducted with the denoising results of PSNR = 33.2642 and SSIM = 0.9206 on JSRT datasets and PSNR = 35.1086 and SSIM = 0.9328 on LIDC datasets. Compared with the state-of-the-art methods, the superior performance of the proposed method is outstanding.

Projection-Based cascaded U-Net model for MR image reconstruction

Background and ObjectiveBackground and Objective: Recent studies in deep learning reveal that the U-Net stands out among the diverse set of deep models as an effective network structure, especially for imaging inverse problems. Initially, the U-Net model was developed to solve segmentation problems for biomedical images while using an annotated dataset. In this paper, we will study a novel application of the U-Net structure for the important inverse problem of MRI reconstruction. Deep networks are particularly efficient for the speed-up of the MR image reconstruction process by decreasing the data acquisition time, and they can significantly reduce the aliasing artifacts caused by the undersampling in the k-space. Our aim is to develop a novel and efficient cascaded U-Net framework for reconstructing MR images from undersampled k-space data. The new framework should have improved reconstruction performance when compared to competing methodologies. MethodsIn this paper, a novel cascaded framework utilizing the U-Net as a sub-block is being proposed. The introduced U-Net cascade structure is applied to the magnetic resonance image reconstruction problem. The connection between the cascaded U-Nets is realized in the form of a recently developed projection-based updated data consistency layer. The novel structure is implemented in the PyTorch environment, which is one of the standards for deep learning implementations. The recently created fastMRI dataset which forms an important benchmark for MRI reconstruction is used for training and testing purposes. ResultsWe present simulation results comparing the novel method with a variety of competitive deep networks. The new cascaded U-Net structures PSNR performance stands on average 1.28 dB higher than the baseline U-Net. The improvement, when compared to the standard CNN, is on average 3.32 dB. ConclusionsThe proposed cascaded U-Net configuration results in an improved reconstruction performance when compared to the CNN, the cascaded CNN, and also the singular U-Net structures, where the singular U-Net forms the baseline reconstruction method from the fastMRI package. The use of the projection-based updated data consistency layer also leads to improved quantitative (including SSIM, PSNR, and NMSE results) and qualitative results when compared to the use of the conventional data consistency layer.

Dynamically Adaptive Fast Motion Estimation Algorithm for HD Video

This paper presents a novel adaptive Fast Motion Estimation (FME) Algorithm, which reduces the number of search points, computational complexity and therefore lowering power consumption, while providing improved quality per watt in FME. The algorithm minimizes computation and storage and from there, power consumption, while trying to preserve quality as much as possible. It is very suitable for hardware implementation, as in a motion co-processor. The pervasiveness of mobile devices with HD video capabilities demands such low power hardware accelerators. The proposed algorithm is a novel method, that can identify the best search pattern within a given region of the HD video frame, based on its motion dynamics to achieve lower power video encoding. The goal is to achieve the best quality with the minimum number of search iterations. The reduced number of checks translates into power savings. The motivation of the presented algorithm is to learn the motion dynamics within a region from a small subset of blocks within that region and apply that knowledge to the remaining large set of blocks in the same region. This adaptiveness makes it robust with respect to the nature of the video. Thus it works equally well with more dynamic and with less dynamic video sequences. The results show that the proposed algorithm can reduce computations by about 4 times compared to other fixed search patterns algorithms, while staying within 1 dB of PSNR results. That equates to about 75% power savings at the expense of not more than 1 dB of PSNR quality loss.

PENERAPAN METODE FILTER GAUSSIAN DALAM PERBAIKAN KUALITAS CITRA SATELIT MODIS

One technology that uses satellite capabilities that can capture the image of an area is the MODIS (Moderate Resolution Imaging Spectroradiometer) sensor, the resulting image may experience noise (noise) this is caused when sending through the transmission line, the image is too dark or bright, the image less sharp or blurred. For that we need a method of noise reduction such as remote sensing through satellite imagery. The purpose of this study is to apply the Gaussian lowpass filter method in improving the quality of fashionable satellite imagery in order to have good image quality. The results of PSNR show the highest value on the MOD_04 image, with smoother image results, the resulting PSNR value is 19.7791 higher than other images, where the higher the PSNR value, the better the resulting image.

Back-projection-based progressive growing generative adversarial network for single image super-resolution

Recent advanced deep learning studies have shown the positive role of feedback mechanism in image super-resolution task. However, current feedback mechanism only calculates residual errors of images with the same resolution without considering the useful features that may be carried by different resolution features. In this paper, to explore the potential of feedback mechanism, we design a new network structure (progressive up- and downsampling back-projection units) to construct a generative adversarial network for single image super-resolution and use progressive growing methodologies to train it. Unlike previous feedback structure, we use progressively increasing scale factor to build up- and down-projection units, which aims to learn fruitful features across scales. This method allows us to get more meaningful information from early feature maps. Additionally, we train our network progressively; in the process of training, we start from single layer network structure and add new layers as the training goes on. By this mean, the training process can be greatly accelerated and stabilized. Experiments on benchmark dataset with the state-of-the-art methods show that our network achieves 0.01 dB, 0.11 dB, 0.13 dB and 0.4 dB better PSNR results than that of RDN+, MDSR, D-DBPN and EDSR on 8 $$\times $$ enlargement, respectively, and also achieves favorable performance against the state-of-the-art methods on 2 $$\times $$ and 4 $$\times $$ enlargement.

Image patch prior learning based on random neighbourhood resampling for image denoising

Image patch priors become a popular tool for image denoising. The Gaussian mixture model (GMM) is remarkably effective in modelling natural image patches. However, GMM prior learning using the expectation maximisation (EM) algorithm is sensitive to the initialisation, often leading to low convergence rate of parameter estimation. In this study, a novel sampling method called random neighbourhood resampling (RNR) is proposed to improve the accuracy and efficiency of parameter estimation. An enhanced GMM (EGMM) learning algorithm is further developed by incorporating RNR into the EM algorithm to initialise and update the GMM prior. The learned EGMM prior is applied in the expected patch log-likelihood (EPLL) framework for image denoising. The effectiveness and performance of the proposed RNR and EGMM algorithm are demonstrated via extensive experimental results comparing with the state-of-the-art image denoising methods, the experimental results show the higher PSNR result of the denoised images using the proposed method. Meanwhile, the authors verified that the proposed method can efficiently reduce the time of image denoising compared with the basic EPLL method.

High Density Video Impulse Noise Reduction Scheme Based on Spatially Growing Modified Median Filter

In this paper a spatially growing modified median filter method has been proposed for the restoration of digital videos which are adulterated by the saturated impulse noise i.e., salt-and pepper noise. The proposed denoising process executes filtering action only to the corrupted pixels in the video keeping noise free pixels in the video unharmed. The current study has used a spatially growing window method for the exclusion of high density noise present in the digital videos. It has used sliding window of increasing dimension centered at any pixel and sequentially swapped the noise corrupted pixels by the median value of the trimmed window. However, if the whole pixels in the window are noise corrupted then the dimension of sliding window is enlarged in order to obtained the noise free pixels for median computing. At the same time, this process has been found to be capable to removing the high density salt and pepper noise and also well preserved the fine details of the videos. Experimentally, it has been found that the proposed method generate enhanced PSNR and SSIM results.

Symbiotic Organisms Search Algorithm for multilevel thresholding of images

Thresholding is a frequently used method in image processing because of its consistency and low computational cost. Otsu's and Kapur's methods are two important techniques that were proved to be best thresholding methods. However, they have high computational complexity when extended to multilevel thresholding because of their exhaustively search. Recently, meta-heuristic algorithms have been successfully applied for thresholding problems. In this study, six different meta-heuristic algorithms based on Otsu's and Kapur's functions; Particle Swarm Optimization (PSO), Firefly Algorithm (FA), Symbiotic Organisms Search (SOS), Artifical Bee Colony (ABC), Genetic Algorithm (GA) and grey Wolf Optimizer (GWO) were used for multilevel thresholding problem and compared. Experimental results suggest that SOS, PSO and FA algorithms often have higher fitness values than other algorithms. Especially when more than two threshold values are determined, SOS algorithm mostly gives higher fitness values. PSNR and SSIM results of the algorithms are similar. In terms of computational complexity, the GWO algorithm has the fastest convergence. For standard deviations of objective functions; more stable results were obtained with SOS based on Kapur's function, SOS and PSO based on Otsu's function. Also, SOS based on Kapur's function was found to be the most successful algorithm in the Friedman test. As a result, although the GWO approached faster, the SOS algorithm produced more consistent results for both objective functions.

Local Foreground Removal Disocclusion Filling Method for View Synthesis

View synthesis is an effective method to generate the contents of multiple views based on a limited number of reference views, which can be used in 2D to 3D conversion, free viewpoint video and multiview video rendering. Depth-image-based rendering (DIBR) is a practical technique to generate virtual view by using a 2D reference view and its depth image. However, a critical problem in DIBR process is that disocclusions might be produced in the synthesized image because the background occluded by the foreground objects in the reference view may be exposed in the virtual view. In this paper, a local foreground removal method is proposed for disocclusion filling. Morphology-based depth image preprocessing is performed before DIBR, aiming to correct the depth value of the ghosts and remove ghost artifacts. In the synthesized virtual image, pixels on the disocclusion edge are identified and classified. Then they are positioned in the reference image by inverse 3D warping. Local foreground regions that occlude the corresponding background are removed from both the reference image and its depth image based on the disocclusion edge pixels. Removed region is filled with surrounding background contents, and depth information is used in this process to prevent foreground penetration. The predicted background contents are warped to the disocclusion region, thereby achieving the hole filling. Experimental results show that the proposed method performs better than the other methods in disocclusion filling, and improves the subjective and objective quality of the synthesized view. In the evaluation results of PSNR, SSIM, FSIMc and VSI, our method improves by 0.32-2.43dB, 0.0036-0.0155, 0.0041-0.0198 and 0.0012-0.0057 respectively compared with competitive methods.

An improved security and message capacity using AES and Huffman coding on image steganography

Information security is very important and has been widely implemented. Cryptography and steganography are two common methods that can be implemented to secure and conceal the information. In this research, the proposed AES algorithm for cryptography and DWT for steganography. However, in case of implementing DWT as steganography, there is a weakness which is a lower capacity. Based on DWT’s problem, proposed Huffman Coding to reduce the total of the message’s bit and increase the capacity. In the implementation, a message will be processed by using AES and compressed by using Huffman Coding then conceal in a cover using DWT. After doing several experiments using a 128x128 pixel message image and a 512x512 pixel of the cover image, achieved the average of MSE is 1.5676 and the average of PSNR result is above 40db which is 46.1878.

Data embedding in high efficiency video coding (HEVC) videos by modifying the partitioning of coding units

A data embedding solution in HEVC videos is proposed by modifying the partitioning of coding units (CUs). The partitions of a CU are first represented as a sequence of binary flags. The flags pertaining to 16 × 16 sub CUs are used as a cover for data embedding, where 6 or 4 message bits are embedded per CU. The data embedding algorithm guarantees that a maximum of one partition is modified per message segment, therefore, in a given CU, either 0, 1 or 2 partitions are modified. The proposed solution is assessed in terms of message payload, number of modified partitions, loss in video quality as indicated by the PSNR results, mean objective scores and excessive bitrate. The proposed solution can embed messages with up to an average payload of 32.6 kbit/s with a corresponding average distortion of <0.5 dB. Comparisons with existing solutions reveal that the proposed solution maintains similar message payloads with less modifications of CU partitioning and at the same time resulting in less distortions for the cover video.

Gaussian Mixture Model and Lifting Wavelet Transformed Base Satellite Image Enhancement

From the last few decades, Satellite images are being used widely in various applications like monitoring of forest areas, weather forecasting, polar bears counting, etc. In those applications to get more details of images efficiently, satellite images should be enhanced up to the required level as the images captured by the satellites are covered very large areas and those are very low-resolution images due to the high altitudes of satellites from the earth. We proposed a method of an image enhancement which includes both resolution enhancement and contrast enhancement. In this method, Stationary Wavelet Transform (SWT) in combination with Lifting Wavelet Transform (LWT) is used for image decomposition into low-frequency sub band images and high-frequency sub band images to separate smooth regions and sharp edges to interpolate regions and edges separately to reduce blurring effect in edges and noise in smooth regions. To get smoother details and sharper edges, Gaussian Mixture Model (GMM) is used for interpolation in resolution enhancement process and SWT with the combination of Contrasts Limited Adaptive Histogram Equalization (CLAHE) for contrast enhancement process. SWT in combination with LWT improves the resolution effectively and also minimizes the execution time drastically than existing methods due to the shift invariance of SWT and reduced computations in LWT and GMM interpolation results from sharper edges and smoother details. SWT is used in combination with CLAHE to enhance the contrast and mitigate the noise effects than existing methods. The proposed method gives superior results and compared with existing techniques with PSNR, Noise Estimation, and visual results.

Teknik Penyembunyian dan Enkripsi Pesan pada Citra Digital dengan Kombinasi Metode LSB dan RSA

This study proposed a combination of steganography and cryptography techniques using LSB and RSA methods. RSA is a popular cryptographic technique that can be applied to digital imagery. Digital image pixel values range from 0 to 255, making the keys used in RSA limited enough and less secure. So in this study, it is proposed to convert pixel image value to 16 bits so that the key used can be more varied. Experimental results proved that there was a steady increase in security and imperceptibility. This was shown by the results of PSNR 57.2258dB and MSE 0.1232dB. This method was also resistant to salt and pepper attacks.

Blind image deblurring via gradient orientation-based clustered coupled sparse dictionaries

In this paper, we proposed a novel sparse representation-based blind image deblurring algorithm, which exploits the benefits of coupled sparse dictionary, and patch gradient orientation-based sparsifying sub-dictionary learning. We jointly trained coupled dictionaries for blurred and clear image patches to take advantages of the similarity of sparse representation in the blurred and clear image patch pair with respect to their corresponding dictionaries. The first step of the algorithm is to estimate blur kernel from the test image itself which is utilized in generating blur image training set from the clear image training set. Instead of learning a large coupled dictionary, we have proposed to cluster the patches having similar geometric structures and learn smaller sub-dictionaries for each group to improve the effectiveness of sparse modeling of the information in an image. While reconstructing the image, the sparse representation of a blurred image patch is applied to the blur-free dictionary to generate a blur-free image patch. For choosing a sub-dictionary which best describes a particular patch, minimum residue error criterion is formulated. An iterative error compensation mechanism is carried out to enhance the deblurring performance and to compensate for sparse approximation. The performance of proposed deblurring method is evaluated in terms of PSNR, SSIM, ISNR, and visual quality results. The simulation results demonstrate that the proposed method achieves very competitive deblurring performance as compared to other complementary blind deblurring methods.

An Adaptive Scheme to Achieve Fine Grained Video Scaling

&lt;p&gt;A robust Adaptive Reconstruction Error Minimization Convolution Neural Network (&lt;strong&gt; ARemCNN&lt;/strong&gt;) architecture introduced to provide high reconstruction quality from low resolution using parallel configuration. Our proposed model can easily train the bulky datasets such as YUV21 and Videoset4.Our experimental results shows that our model outperforms many existing techniques in terms of PSNR, SSIM and reconstruction quality. The experimental results shows that our average PSNR result is 39.81 considering upscale-2, 35.56 for upscale-3 and 33.77 for upscale-4 for Videoset4 dataset which is very high in contrast to other existing techniques. Similarly, the experimental results shows that our average PSNR result is 38.71 considering upscale-2, 34.58 for upscale-3 and 33.047 for upscale-4 for YUV21 dataset.&lt;/p&gt;

REAL-TIME VIDEO SCALING BASED ON CONVOLUTION NEURAL NETWORK ARCHITECTURE

In recent years, video super resolution techniques becomes mandatory requirements to get high resolution videos. Many super resolution techniques researched but still video super resolution or scaling is a vital challenge. In this paper, we have presented a real-time video scaling based on convolution neural network architecture to eliminate the blurriness in the images and video frames and to provide better reconstruction quality while scaling of large datasets from lower resolution frames to high resolution frames. We compare our outcomes with multiple exiting algorithms. Our extensive results of proposed technique RemCNN (Reconstruction error minimization Convolution Neural Network) shows that our model outperforms the existing technologies such as bicubic, bilinear, MCResNet and provide better reconstructed motioning images and video frames. The experimental results shows that our average PSNR result is 47.80474 considering upscale-2, 41.70209 for upscale-3 and 36.24503 for upscale-4 for Myanmar dataset which is very high in contrast to other existing techniques. This results proves our proposed model real-time video scaling based on convolution neural network architecture's high efficiency and better performance.

Real-Time Video Scaling Based on Convolution Neural Network Architecture

In recent years, video super resolution techniques becomes mandatory requirements to get high resolution videos. Many super resolution techniques researched but still video super resolution or scaling is a vital challenge. In this paper, we have presented a real-time video scaling based on convolution neural network architecture to eliminate the blurriness in the images and video frames and to provide better reconstruction quality while scaling of large datasets from lower resolution frames to high resolution frames. We compare our outcomes with multiple exiting algorithms. Our extensive results of proposed technique RemCNN (Reconstruction error minimization Convolution Neural Network) shows that our model outperforms the existing technologies such as bicubic, bilinear, MCResNet and provide better reconstructed motioning images and video frames. The experimental results shows that our average PSNR result is 47.80474 considering upscale-2, 41.70209 for upscale-3 and 36.24503 for upscale-4 for Myanmar dataset which is very high in contrast to other existing techniques. This results proves our proposed model real-time video scaling based on convolution neural network architecture’s high efficiency and better performance.

Visible and Infrared Data Fusion

This paper shows data fusion method with visible light image and infrared image. We assume an image is captured by FLIR image, which provides two images: visible light and infrared images. The details of the image are obtained in visible light image. This information is added to original low resolution infrared image and we assume this image as the detail strengthen infrared image. Experimental results section provides PSNR and MSE results. It is obvious that that proposed method with p=0.03 gives visually satisfactory results.

Image encryption based on Independent Component Analysis and Arnold’s Cat Map

Security of the multimedia data including image and video is one of the basic requirements for the telecommunications and computer networks. In this paper, a new efficient image encryption technique is presented. It is based on modifying the mixing matrix in Independent Component Analysis (ICA) using the chaotic Arnold’s Cat Map (ACM) for encryption. First, the mixing matrix is generated from the ACM by insert square image of any dimension. Second, the mixing process is implemented using the mixing matrix and the image sources the result is the encryption images that depend on the number of sources. Third, images decrypted using ICA algorithms. We use the Joint Approximate Diagonalization of Eigen-matrices (JADE) algorithm as a case study. The results of several experiments, PSNR, SDR and SSIM index tests compared with standard mixing matrix showed that the proposed image encryption system provided effective and safe way for image encryption.

Comparison Between Various Standard Definition to High Definition Image Conversion Methods

Different methods of standard definition (SD) to high definition (HD) image conversion are compared according to the PSNR criterion. Subjective comparison is difficult as there is almost no quality difference between them. For comparison, an original HD and the correspondent SD images are required. The SD image is up-converted to approximate HD images using various methods and then these approximated images are compared to the original HD image. In such a comparison the main question is how to generate the SD image. Both analytical and simulation results show that the method of generation of the SD image has a great influence on the outcome of PSNR results. Immediate down-sampling, Filtered down-sampling, Discrete Fourier Transform (DFT)-based down-sampling and Discrete Cosine Transform (DCT)-based down-sampling are methods we used to generate SD image from the original HD image. Ideal low pass signal (Nyquist) interpolation, DFTbased up-sampling and DCT-based up-sampling are the three methods we used to up-convert SD to HD image. As it is seen, the highest PSNR (least error) is the mean squared value of the omitted high frequency components during down-sampling process in transform-based methods. The analytical formulation of PSNR in filtered-based methods is not simple due to distortions in the frequency response of filters and aliasing in frequency domain. Keywords: Discrete cosine transform, discrete fourier transform, down-sampling, nyquist rate, PSNR, upsampling.