Deep Convolutional Autoencoders Research Articles

Automatic Facial Expression Recognition Using Modified LPQ and HOG Features with Stacked Deep Convolutional Autoencoders

Automatic Facial Expression Recognition Using Modified LPQ and HOG Features with Stacked Deep Convolutional Autoencoders

ENHANCED SECURITY MEASURES FOR NATIONAL IDENTITY CARD USING IMAGE STEGANOGRAPHY

National identity documents are essential cards issued by official authorities, typically containing a photo and used for various purposes within a country, such as travel, electronic identification, and access to secure locations. These documents incorporate multiple security features to combat forgery, yet criminals increasingly target the manipulation of genuine documents and facial images. Trusted identity is crucial for societal function, necessitating continuous improvements in security measures by governments and ID manufacturers. To address this, we present StegoCard, a novel steganography method specifically designed for concealing messages within facial images commonly found on IDs. StegoCard employs a series of Deep Convolutional Auto Encoders to embed secret messages into facial portraits, creating stego facial images. These stego images can then be decoded by Deep Convolutional Auto Decoders, even after being printed and captured digitally. Our StegoCard approach outperforms existing methods like StegaStamp in terms of perceptual quality, as demonstrated by metrics such as Peak Signal-to-Noise Ratio, hiding capacity, and imperceptibility. By leveraging deep learning and steganography, StegoCard enhances the security of national identity documents, contributing to the preservation of trusted identities within society. KEYWORD:Deep Learning, Steganography,Recurrrent proposal Network(RPN)

Convolutional Autoencoder-Based medical image compression using a novel annotated medical X-ray imaging dataset

With the ongoing development of machine learning techniques, it is now necessary to train and evaluate these algorithms to have access to high-quality medical X-ray datasets. This study unfolds on two critical axes within the realm of medical imaging. We introduce the proposed Medical X-ray Imaging Dataset (MXID), a meticulously curated resource featuring images spanning 18 body parts. It has been refined to include a classification based on body type within each gender category. This dataset addresses the limitations of existing datasets by offering comprehensive coverage, precise annotations, and optimal image quality. While numerous datasets often fall short in terms of encompassing a broad range of anatomical regions, particularly those that are designed for multi-body component analysis. Our primary contribution lies in bridging the gap in available resources, providing a foundation of developing robust machine learning algorithms in medical imaging. Moreover, we extend our investigation to the integration of the MXID dataset into the domain of medical image compression, evaluating the performance of techniques such as Principal Component Analysis (PCA), K-means clustering, Convolutional Neural Network (CNN), Deep Convolutional Autoencoders (DCAEs), Autoencoders (AEs), and Variational Autoencoder (VAE), using two scenarios, one is for single image compression, and the other is for MXID dataset, which offer a unique perspective that enables a comprehensive assessment of these techniques across diverse anatomical regions. Our study contributes to advancing machine learning and deep learning applications in medical imaging, emphasizing the creation of a dataset and its integration into image compression.

Deep reinforcement learning and convolutional autoencoders for anomaly detection of congenital inner ear malformations in clinical CT images

Detection of abnormalities within the inner ear is a challenging task even for experienced clinicians. In this study, we propose an automated method for automatic abnormality detection to provide support for the diagnosis and clinical management of various otological disorders. We propose a framework for inner ear abnormality detection based on deep reinforcement learning for landmark detection which is trained uniquely in normative data. In our approach, we derive two abnormality measurements: Dimage and Uimage. The first measurement, Dimage, is based on the variability of the predicted configuration of a well-defined set of landmarks in a subspace formed by the point distribution model of the location of those landmarks in normative data. We create this subspace using Procrustes shape alignment and Principal Component Analysis projection. The second measurement, Uimage, represents the degree of hesitation of the agents when approaching the final location of the landmarks and is based on the distribution of the predicted Q-values of the model for the last ten states. Finally, we unify these measurements in a combined anomaly measurement called Cimage. We compare our method’s performance with a 3D convolutional autoencoder technique for abnormality detection using the patch-based mean squared error between the original and the generated image as a basis for classifying abnormal versus normal anatomies. We compare both approaches and show that our method, based on deep reinforcement learning, shows better detection performance for abnormal anatomies on both an artificial and a real clinical CT dataset of various inner ear malformations with an increase of 11.2% of the area under the ROC curve. Our method also shows more robustness against the heterogeneous quality of the images in our dataset.

Multiview Deep Subspace Clustering Networks.

Multiview subspace clustering aims to discover the inherent structure of data by fusing multiple views of complementary information. Most existing methods first extract multiple types of handcrafted features and then learn a joint affinity matrix for clustering. The disadvantage of this approach lies in two aspects: 1) multiview relations are not embedded into feature learning and 2) the end-to-end learning manner of deep learning is not suitable for multiview clustering. Even when deep features have been extracted, it is a nontrivial problem to choose a proper backbone for clustering on different datasets. To address these issues, we propose the multiview deep subspace clustering networks (MvDSCNs), which learns a multiview self-representation matrix in an end-to-end manner. The MvDSCN consists of two subnetworks, i.e., a diversity network (Dnet) and a universality network (Unet). A latent space is built using deep convolutional autoencoders, and a self-representation matrix is learned in the latent space using a fully connected layer. Dnet learns view-specific self-representation matrices, whereas Unet learns a common self-representation matrix for all views. To exploit the complementarity of multiview representations, the Hilbert-Schmidt independence criterion (HSIC) is introduced as a diversity regularizer that captures the nonlinear, high-order interview relations. Because different views share the same label space, the self-representation matrices of each view are aligned to the common one by universality regularization. The MvDSCN also unifies multiple backbones to boost clustering performance and avoid the need for model selection. Experiments demonstrate the superiority of the MvDSCN.

Deep convolutional autoencoders for the time–space reconstruction of liquid rocket engine flames

The integration of high-fidelity numerical simulations into the rocket engine design-cycle promises to cut costs in an ever more competitive launch market. Although still being intractable in the optimization of most of combustion devices, their potential can be harnessed via appropriate reduced order models (ROMs). In recent years, significant progress has been made in the ROM literature through the adoption of non-linear methods, coming from the realm of deep learning. These methods promise to deliver where conventional linear techniques, such as Proper Orthogonal Decomposition (POD), fall short due to their inherent qualities.In the current work, we carry out a first attempt to develop purely data-driven emulators for a shear-coaxial liquid rocket engine injector. In particular, a Convolutional Neural Network autoencoder (CNN-AE) with a Multi-Layer Perceptron (MLP) latent dynamics encoder architecture is considered. The main objective is to assess the capacity of autoencoders based models to reconstruct, in time and space, the instantaneous temperature field of a non-canonical, complex, reactive, flow configuration. The data used is made of Large Eddy Simulations (LES) of three distinct, real-life inspired, injector configurations.The autoencoder based models show superior reconstruction capabilities compared to POD with similar latent-space dimension. Moreover, the time–space reconstructed fields preserve the spectral content of the snapshots, with only a small loss in gain at high frequencies. These results serve as proof-of-concept for further developments on surrogate models of liquid rocket engine injectors.

AI‐enhanced iterative solvers for accelerating the solution of large‐scale parametrized systems

SummaryRecent advances in the field of machine learning open a new era in high performance computing for challenging computational science and engineering applications. In this framework, the use of advanced machine learning algorithms for the development of accurate and cost‐efficient surrogate models of complex physical processes has already attracted major attention from scientists. However, despite their powerful approximation capabilities, surrogate model predictions are still far from being near to the ‘exact’ solution of the problem. To address this issue, the present work proposes the use of up‐to‐date machine learning tools in order to equip a new generation of iterative solvers of linear equation systems, capable of very efficiently solving large‐scale parametrized problems at any desired level of accuracy. The proposed approach consists of the following two steps. At first, a reduced set of model evaluations is performed using a standard finite element methodology and the corresponding solutions are used to establish an approximate mapping from the problem's parametric space to its solution space using a combination of deep feedforward neural networks and convolutional autoencoders. This mapping serves as a means of obtaining very accurate initial predictions of the system's response to new query points at negligible computational cost. Subsequently, an iterative solver inspired by the Algebraic Multigrid method in combination with Proper Orthogonal Decomposition, termed POD‐2G, is developed that successively refines the initial predictions of the surrogate model towards the exact solution. The application of POD‐2G as a standalone solver or as preconditioner in the context of preconditioned conjugate gradient methods is demonstrated on several numerical examples of large scale systems, with the results indicating its strong superiority over conventional iterative solution schemes.

Stellar Karaoke: deep blind separation of terrestrial atmospheric effects out of stellar spectra by velocity whitening

ABSTRACT We report a study exploring how the use of deep neural networks with astronomical Big Data may help us find and uncover new insights into underlying phenomena: through our experiments towards unsupervised knowledge extraction from astronomical Big Data we serendipitously found that deep convolutional autoencoders tend to reject telluric lines in stellar spectra. With further experiments, we found that only when the spectra are in the barycentric frame does the network automatically identify the statistical independence between two components, stellar versus telluric, and rejects the latter. We exploit this finding and turn it into a proof-of-concept method for removal of the telluric lines from stellar spectra in a fully unsupervised fashion: we increase the interobservation entropy of telluric absorption lines by imposing a random, virtual radial velocity to the observed spectrum. This technique results in a non-standard form of ‘whitening’ in the atmospheric components of the spectrum, decorrelating them across multiple observations. We process more than 250 000 spectra from the High Accuracy Radial velocity Planetary Search and with qualitative and quantitative evaluations against a data base of known telluric lines, show that most of the telluric lines are successfully rejected. Our approach, ‘Stellar Karaoke’, has zero need for prior knowledge about parameters such as observation time, location, or the distribution of atmospheric molecules and processes each spectrum in milliseconds. We also train and test on Sloan Digital Sky Survey and see a significant performance drop due to the low resolution. We discuss directions for developing tools on top of the introduced method in the future.

Fake Face Detection in Identity Cards Using Stegoface

Abstract: Identification and Machine-Readable Travel Documents (MRTDs) play a crucial role in verifying and validating identities in various situations, such as international travel, civil applications, online commerce, and access to transaction processing systems. These documents incorporate multiple security features aimed at preventing document forgery. However, criminals have shifted their focus to obtaining genuine documents fraudulently and manipulating facial portraits, as the existing security systems are challenging to bypass. To address this issue and mitigate the risks associated with such fraud, it is imperative for governments and ID/MRTD manufacturers to continually enhance and develop security measures. In this context, we present StegoFace, an innovative and efficient steganography method specifically designed for concealing secret messages within facial images found on common IDs and MRTDs. StegoFace employs an end-to-end approach, consisting of an ensemble of n Deep Convolutional Auto Encoders, which encode the secret message into a stegofacial image, and a Deep Convolutional Auto Decoder, capable of extracting the hidden message from the stegofacial image. Notably, our StegoFace approach outperforms the StegaStamp method in terms of perceptual quality, as evidenced by the results of metrics such as Peak Signal-to-Noise Ratio, hiding capacity, and imperceptibility on the test dataset.

In situ grape ripeness estimation via hyperspectral imaging and deep autoencoders

The estimation of the grapes’ maturity in the field using non-destructive techniques is of high interest for the high-valued vinified grapes, particularly towards the development of fully automated agrobots that perform selective harvesting operations. Whereas infrared spectroscopy has been employed using point spectrometers in the laboratory and in the field, imaging spectrometers have mainly been tested in controlled laboratory conditions due to issues with varying illumination. In this paper, the application of the autoencoder framework is proposed, which is employed to transform the raw recorded spectra, regardless of illumination conditions, into standardized reflectance spectra; thus addressing the inherent difficulties which hamper the direct application of hyperspectral imaging in the field. To validate the methodology, the sugar content (∘Brix) of four grape varieties, namely Chardonnay, Malagouzia, Sauvignon-Blanc, and Syrah, is estimated. Two different autoencoder architectures are examined: deep fully-connected (DAE) and deep convolutional autoencoders (DCAE), while the estimation of sugar content takes place using as input both from the encoded (latent) space and from the autoencoders’ output, i.e., the transformed standardized spectra. The use of multiple spectral pre-treatments is further examined to enhance the accuracy of prediction. Despite that DAE and DCAE showcase comparable similarity metrics, DCAE statistically outperforms DAE when using both the encoded space and the autoencoders’ output, attesting to the suitability of the convolutional autoencoder framework. On the other hand, there is no statistical significant difference when employing multiple input pre-treatments. The accuracy of estimation (mean RMSE 1.83 ∘Brix, R2 0.70, RPIQ 2.43) is comparable to other studies that directly work with standardized reflectance spectra in laboratory conditions.

Reduced-order modeling for stochastic large-scale and time-dependent flow problems using deep spatial and temporal convolutional autoencoders

A non-intrusive reduced-order model based on convolutional autoencoders is proposed as a data-driven tool to build an efficient nonlinear reduced-order model for stochastic spatiotemporal large-scale flow problems. The objective is to perform accurate and rapid uncertainty analyses of the flow outputs of interest for which the input parameters are deemed uncertain. The data are constituted from a set of high-fidelity snapshots, collected using an inhouse high-fidelity flow solver, which correspond to a sample of the uncertain input parameters. The method uses a 1D-convolutional autoencoder to reduce the spatial dimension of the unstructured meshes used by the flow solver. Another convolutional autoencoder is used for the time compression. The encoded latent vectors, generated from the two compression levels, are then mapped to the input parameters using a regression-based multilayer perceptron. The proposed model allows for rapid predictions for unseen parameter values, allowing the output statistical moments to be computed efficiently. The accuracy of the proposed approach is compared to that of the linear reduced-order technique based on an artificial neural network through two benchmark tests (the one-dimensional Burgers and Stoker’s solutions) and a hypothetical dam break flow problem, with an unstructured mesh and over a complex bathymetry river. The numerical results show that the proposed methods present strong predictive capabilities to accurately approximate the statistical moments of the outputs. In particular, the predicted statistical moments are oscillations-free, unlike those obtained with the traditional proper orthogonal decomposition method. The proposed reduction framework is simple to implement and can be applied to other parametric and time-dependent problems governed by partial differential equations, which are commonly encountered in many engineering and science problems.

Image-Based Detection of Modifications in Assembled PCBs with Deep Convolutional Autoencoders.

In this paper, we introduce a one-class learning approach for detecting modifications in assembled printed circuit boards (PCBs) based on photographs taken without tight control over perspective and illumination conditions. Anomaly detection and segmentation are essential for several applications, where collecting anomalous samples for supervised training is infeasible. Given the uncontrolled environment and the huge number of possible modifications, we address the problem as a case of anomaly detection, proposing an approach that is directed towards the characteristics of that scenario, while being well suited for other similar applications. We propose a loss function that can be used to train a deep convolutional autoencoder based only on images of the unmodified board-which allows overcoming the challenge of producing a representative set of samples containing anomalies for supervised learning. We also propose a function that explores higher-level features for comparing the input image and the reconstruction produced by the autoencoder, allowing the segmentation of structures and components that differ between them. Experiments performed on a dataset built to represent real-world situations (which we made publicly available) show that our approach outperforms other state-of-the-art approaches for anomaly segmentation in the considered scenario, while producing comparable results on a more general object anomaly detection task.

Evaluating different deep learning models for efficient extraction of Raman signals from CARS spectra.

The nonresonant background (NRB) contribution to the coherent anti-Stokes Raman scattering (CARS) signal distorts the spectral line shapes and thus degrades the chemical information. Hence, finding an effective approach for removing NRB and extracting resonant vibrational signals is a challenging task. In this work, a bidirectional LSTM (Bi-LSTM) neural network is explored for the first time to remove the NRB in the CARS spectra automatically, and the results are compared with those of three DL models reported in the literature, namely, convolutional neural network (CNN), long short-term memory (LSTM) neural network, and very deep convolutional autoencoders (VECTOR). The results of the synthetic test data have shown that the Bi-LSTM model accurately extracts the spectral lines throughout the range. In contrast, the other three models' efficiency deteriorated while predicting the peaks on either end of the spectra, which resulted in a 60 times higher mean square error than that of the Bi-LSTM model. The Pearson correlation analysis demonstrated that Bi-LSTM model performance stands out from the rest, where 94% of the test spectra have correlation coefficients of more than 0.99. Finally, these four models were evaluated on four complex experimental CARS spectra, namely, protein, yeast, DMPC, and ADP, where the Bi-LSTM model has shown superior performance, followed by CNN, VECTOR, and LSTM. This comprehensive study provides a giant leap toward simplifying the analysis of complex CARS spectroscopy and microscopy.

Unsupervised Clustering of Microseismic Signals Using a Contrastive Learning Model

Distinguishing useful microseismic signals is a critical step in microseismic monitoring. Here, we present the time series contrastive clustering (TSCC) method, an end-to-end unsupervised model for clustering microseismic signals that employs a contrastive learning network and a centroidal-based clustering model. The TSCC framework consists of two successive phases: pre-training and fine-tuning. In the pre-training phase, two random cropping augmentations are used to transform the time series microseismic data into two distinct but correlated views. Then, the multi-scale temporal and instance contrasting learning are used to discriminate between negative and positive views, thus motivating the encoder to capture microseismic signal contextual information from multiple perspectives and generate distinct representations from unlabeled data. During the fine-tuning phase, the encoder weights are iteratively fine-tuned by simultaneously performing contrast learning and clustering. The corresponding loss is a weighted combination of the contrastive and clustering loss functions, which induces the encoder to learn representations that improve clustering performance. The test results demonstrate that the proposed method can achieve better clustering accuracy than popular clustering methods, including k-means, deep embedding clustering, unsupervised clustering with deep convolutional autoencoders, and deep clustering with self-supervision. Moreover, the TSCC model can produce results comparable to supervised deep learning approaches while requiring no labeled data, manual feature extraction, or large training datasets. In practice, the TSCC model has a clustering accuracy of 98.07% and a normalized mutual information of 86.26%.

Deep convolutional auto-encoders for emotion recognition from facial expressions

One of the unsolved problems in computer vision is recognizing or understanding other people's emotions and feelings. Deep Convolutional Neural Networks (CNNs) have attempted to be effective in addressing emotion recognition issues. The significant level of performance achieved by these classifiers can be attributed to their ability to self-learn a downsampled feature vector that retains abstraction information through filter kernels in convolutional layers. Despite these advancements, challenges persist in capturing subtle and context-dependent emotional nuances, hindering the development of more nuanced emotion recognition systems. Ongoing research explores multimodal approaches, integrating various sensory inputs, to enhance the accuracy and reliability of emotion detection in real-world scenarios. In this paper, we explore the impact of training the initial weights in an unsupervised manner. We study the results of pretraining a Deep CNN as a Convolutional Auto-Multiplexer (CAM) in a greedy layer-wise unsupervised fashion for emotion recognition using facial feature images. When trained with randomly initialized weights, our CNN emotion recognition model achieves a performance rate of 92.16% on the Karolinska Directed Emotional Faces (KDEF) dataset. In contrast, by using this pre-trained model, the performance increases to 93.52%. Pre-training our CNN as a CAM also marginally reduces training time.

Pseudo-Supervised Defect Detection Using Robust Deep Convolutional Autoencoders

Robust Autoencoders separate the input image into a Signal(L) and a Noise(S) part which, intuitively speaking, roughly corresponds to a more stable background scene (L) and an undesired anomaly (or defect) (S). This property of the method provides a convenient theoretical basis for divorcing intermittent anomalies that happen to clutter a relatively consistent background image. In this paper, we illustrate the use of Robust Deep Convolutional Autoencoders (RDCAE) for defect detection, via a pseudo-supervised training process. Our method introduces synthetic simulated defects (or structured noise) to the training process, that alleviates the scarcity of true (real-life) anomalous samples. As such, we offer a pseudo-supervised training process to devise a well-defined mechanism for deciding that the defect-normal discrimination capability of the autoencoders has reached to an acceptable point at training time. The experiment results illustrate that pseudo supervised Robust Deep Convolutional Autoencoders are very effective in identifying surface defects in an efficient way, compared to state of the art anomaly detection methods.

Dissimilate-and-assimilate strategy for video anomaly detection and localization

Unsupervised anomaly detection in videos is a challenging task owing to the remarkable generalization capacity of the deep convolutional autoencoders and the complex nature of anomalous events. In this study, we introduce a dissimilate-and-assimilate strategy to learn essential patterns of multilevel latent representations of normal spatial and temporal information. To obtain the core normality of the appearance and motion samples over multiple layers of the network, our proposed method diversifies the latent patterns of normal spatial and temporal data to make the out-of-distribution samples discrete (dissimilation) and integrates the latent features of two different samples into a single sample using a feature attention mechanism for robust optimization (assimilation). Based on the learned representations, the network generates convincing predictions of the normal frame, even if it receives abnormal samples after training. That is, the anomalous objects in a series of frames can be detected with significant reconstruction errors, thus leading to better detection and precise localization performance. To verify the effectiveness of the proposed method, we quantify the preciseness of anomaly localization using the outside-inside error ratio along with the traditional area under the curve (AUC) metric to measure the detection performance on the USCD Pedestrian 2, CHUK Avenue and ShanghaiTech Campus datasets.

Reduced order modeling for flow and transport problems with Barlow Twins self-supervised learning

We propose a unified data-driven reduced order model (ROM) that bridges the performance gap between linear and nonlinear manifold approaches. Deep learning ROM (DL-ROM) using deep-convolutional autoencoders (DC–AE) has been shown to capture nonlinear solution manifolds but fails to perform adequately when linear subspace approaches such as proper orthogonal decomposition (POD) would be optimal. Besides, most DL-ROM models rely on convolutional layers, which might limit its application to only a structured mesh. The proposed framework in this study relies on the combination of an autoencoder (AE) and Barlow Twins (BT) self-supervised learning, where BT maximizes the information content of the embedding with the latent space through a joint embedding architecture. Through a series of benchmark problems of natural convection in porous media, BT–AE performs better than the previous DL-ROM framework by providing comparable results to POD-based approaches for problems where the solution lies within a linear subspace as well as DL-ROM autoencoder-based techniques where the solution lies on a nonlinear manifold; consequently, bridges the gap between linear and nonlinear reduced manifolds. We illustrate that a proficient construction of the latent space is key to achieving these results, enabling us to map these latent spaces using regression models. The proposed framework achieves a relative error of 2% on average and 12% in the worst-case scenario (i.e., the training data is small, but the parameter space is large.). We also show that our framework provides a speed-up of 7 times 10^{6} times, in the best case, and 7 times 10^{3} times on average compared to a finite element solver. Furthermore, this BT–AE framework can operate on unstructured meshes, which provides flexibility in its application to standard numerical solvers, on-site measurements, experimental data, or a combination of these sources.

Unsupervised segmentation of biomedical hyperspectral image data: tackling high dimensionality with convolutional autoencoders.

Information about the structure and composition of biopsy specimens can assist in disease monitoring and diagnosis. In principle, this can be acquired from Raman and infrared (IR) hyperspectral images (HSIs) that encode information about how a sample's constituent molecules are arranged in space. Each tissue section/component is defined by a unique combination of spatial and spectral features, but given the high dimensionality of HSI datasets, extracting and utilising them to segment images is non-trivial. Here, we show how networks based on deep convolutional autoencoders (CAEs) can perform this task in an end-to-end fashion by first detecting and compressing relevant features from patches of the HSI into low-dimensional latent vectors, and then performing a clustering step that groups patches containing similar spatio-spectral features together. We showcase the advantages of using this end-to-end spatio-spectral segmentation approach compared to i) the same spatio-spectral technique not trained in an end-to-end manner, and ii) a method that only utilises spectral features (spectral k-means) using simulated HSIs of porcine tissue as test examples. Secondly, we describe the potential advantages/limitations of using three different CAE architectures: a generic 2D CAE, a generic 3D CAE, and a 2D convolutional encoder-decoder architecture inspired by the recently proposed UwU-net that is specialised for extracting features from HSI data. We assess their performance on IR HSIs of real colon samples. We find that all architectures are capable of producing segmentations that show good correspondence with HE stained adjacent tissue slices used as approximate ground truths, indicating the robustness of the CAE-driven spatio-spectral clustering approach for segmenting biomedical HSI data. Additionally, we stress the need for more accurate ground truth information to enable a precise comparison of the advantages offered by each architecture.

Ultrasonic image denoising using machine learning in point contact excitation and detection method

A point contact/Coulomb coupling technique is generally used for visualizing the ultrasonic waves in Lead Zirconate Titanate (PZT) ceramics. The point contact and delta pulse excitation produce a broadband frequency spectrum and wide directional wave vector. In ultrasonic, the signal is corrupted with several types of noises such as speckle, Gaussian, Poisson, and salt and pepper noise. Consequently, the resolution and quality of the images are degraded. The reliability of the health assessment of any civil or mechanical structures highly depends on the ultrasonic signals acquired from the sensors. Recently, deep learning (DL) has been implemented for the reduction of noises from the signals and in images. Here, we have implemented deep learning-based convolutional autoencoders for suitable noise modeling and subsequently denoising the ultrasonic images. Two different metrics, PSNR and SSIM are calculated for quantitative analysis of ultrasonic images. PSNR provides higher visual interpretation, whereas the SSIM can be used to measure much finer similarities. Based upon these parameters speckle-noise demonstrated better than other noise models.