Multilevel Wavelet Transformation Research Articles

Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies

The Reynolds-averaged Navier–Stokes equation for compressible flow over supercritical airfoils under various flow conditions must be rapidly and accurately solved to shorten design cycles for such airfoils. Although deep-learning methods can effectively predict flow fields, the accuracy of these predictions near sensitive regions and their generalizability to large-scale datasets in engineering applications must be enhanced. In this study, a modified vision transformer-based encoder–decoder network is designed for the prediction of transonic flow over supercritical airfoils. In addition, four methods are designed to encode the geometric input with various information points and the performances of these methods are compared. The statistical results show that these methods generate accurate predictions over the complete flow field, with a mean absolute error on the order of 1 × 10−4. To increase accuracy near the shock area, multilevel wavelet transformation and gradient distribution losses are introduced into the loss function. This results in the maximum error that is typically observed near the shock area decreasing by 50%. Furthermore, the models are pretrained through transfer learning on large-scale datasets and fine-tuned on small datasets to improve their generalizability in engineering applications. The results generated by various pretrained models demonstrate that transfer learning yields a comparable accuracy from a reduced training time.

A novel approach towards high-performance image compression using multilevel wavelet transformation for heterogeneous datasets

A novel approach towards high-performance image compression using multilevel wavelet transformation for heterogeneous datasets

Sparse-View CT Reconstruction Based on a Hybrid Domain Model with Multi-Level Wavelet Transform.

The reconstruction of sparsely sampled projection data will generate obvious streaking artifacts, resulting in image quality degradation and affecting medical diagnosis results. Wavelet transform can effectively decompose directional components of image, so the artifact features and edge details with high directionality can be better detected in the wavelet domain. Therefore, a hybrid domain method based on wavelet transform is proposed in this paper for the sparse-view CT reconstruction. The reconstruction model combines wavelet, spatial, and radon domains to restore the projection consistency and enhance image details. In addition, the global distribution of artifacts requires the network to have a large receptive field, so that a multi-level wavelet transform network (MWCNN) is applied to the hybrid domain model. Wavelet transform is used in the encoding part of the network to reduce the size of feature maps instead of pooling operation and inverse wavelet transform is deployed in the decoding part to recover image details. The proposed method can achieve PSNR of 41.049 dB and SSIM of 0.958 with 120 projections of three angular intervals, and obtain the highest values in this paper. Through the results of numerical analysis and reconstructed images, it shows that the hybrid domain method is superior to the single-domain methods. At the same time, the multi-level wavelet transform model is more suitable for CT reconstruction than the single-level wavelet transform.

Automated accurate speech emotion recognition system using twine shuffle pattern and iterative neighborhood component analysis techniques

Speech emotion recognition is one of the challenging research issues in the knowledge-based system and various methods have been recommended to reach high classification capability. In order to achieve high classification performance in speech emotion recognition, a nonlinear multi-level feature generation model is presented by using cryptographic structure. The novelty of this work is the use of cryptographic structure called shuffle box for feature generation and iterative neighborhood component analysis to select the features. The proposed method has three main stages: (i) multi-level feature generation using Tunable Q wavelet transform (TQWT), (ii) twine shuffle pattern (twine-shuf-pat) for feature generation, and (iii) discriminative features are selected using iterative neighborhood component analysis (INCA) and classified. The TQWT is a multi-level wavelet transformation method used to generate high-level, medium-level, and low-level wavelet coefficients. The proposed twine-shuf-pat technique is used to extract the features from the decomposed wavelet coefficients. INCA feature selector is employed to select the clinically significant features. The performance of the obtained model is validated using four speech emotion public databases (RAVDESS Speech, Emo-DB (Berlin), SAVEE, and EMOVO). Our developed twine-shuf-pat and INCA based method yielded 87.43%, 90.09%, 84.79%, and 79.08% classification accuracies using RAVDESS, Emo-DB (Berlin), SAVEE and EMOVO corpora respectively with 10-fold cross-validation strategy. A mixed database is created from four public speech emotion databases which yielded 80.05% classification accuracy. Our obtained speech emotion model is ready to be tested with huge database and can be used in healthcare applications.

Evaluating the Robustness of Image Watermarking System based on Multilevel Wavelet Transform

Evaluating the Robustness of Image Watermarking System based on Multilevel Wavelet Transform

Detection of Breast Cancer using Digital Image Processing Techniques

Breast cancer, a major public health contention and currently causes an increased rate of cancer death in women. The preeminent intent of this project in medical diagnostics is by using mammography, that is a unique imaging technique in medicine for examining the breasts. A higher quality mammographic images (for example electronic pictures) are stored, mammography method (i.e exam) is performed, which is a prior stage for detection and diagnosis of bosom’s malignant growth. In order to detect the tumor FCM algorithm for segmentation is used and the features are extracted by using multi-level wavelet transformation technique with PCA and then some features are added with GLCM features. Further, those segmented region features are extracted and the dataset is trained and tested completely. The images are classified by using SVM, KNN, Tree classifier, Neural Networks or Naive Bayes classifier. Finally, the images from Kaggle dataset are compared and categorized as normal, benign or malignant tumors.

Noise Suppressed Output Voltage Comparison Based on Two Rules of the Mlwt and SC-Ulwt for a PEM Fuel Cell

An environmental crisis due to an indiscreet act of fossil fuels gives rise to the necessity to obtain some renewable solution such as hydrogen fuel cells, geothermal, tidal, solar, biomass, and wind energy. Above all things, an interest in fuel cell has been more increased in academic and industry fields. Among them, the proton exchange membrane (PEM) fuel cell has been recognized as the optimal solution because of having advantages such as low-operating temperature, high potential for low volume and cost, fast start-up ability and high current density, etc. After selection of the PEM fuel cell, the next step is to establish the fuel cell management system (FCMS) for a reliable and an efficient operation. Among several techniques in the FCMS, an equivalent electrical-circuit modeling (EECM) is absolutely indispensable to know fuel cell’s electrochemical and dynamic characteristics. According to more sophisticated EECM, it is possible to predict the fuel cell’s performance in detail. The performance on the EECM is definitely connected with achieving an elaborate terminal voltage. Namely, if the EECM-based estimated- and the measured-voltages are approximately same, it can be judged that the well-designed EECM is accomplished. Nevertheless, it is not easy to obtain an elaborate terminal voltage because unanticipated and instantaneous sensing of noisy may be included in the output voltage. (This output voltage is named as output terminal voltage; OTV). Therefore, it is required to construct the noise suppressed methodology and should be applied as de-noised OTV in the EECM. Little solution for noise suppression methodologies have been known, moreover these have been provided no definitive result. Wavelet transform (WT)-based noise suppression is preferentially selected as the basic methodology in this research. The OVT is considered as the non-stationary signal with different time interval and transient phenomena, and is used as an input signal in the WT. These rules basically implemented the multi-resolution analysis (MRA) that have a vigorous function of both time and frequency localization. Based on this MRA, an original OVT (with noise) is divided into two components such as low- and high-frequency component, An and Dn . Two components have respectively corresponding coefficients, approximate- aj ,k and detailed coefficients dj ,k , for implementation of the decomposition and reconstruction processes in the MRA. Among two coefficients, because of abrupt change during a short period of time of noise, the detailed coefficient dj ,k should be adjusted to suppress noise included in an original OVT. In general, the performance on noise suppression is intimately linked with decomposition and reconstruction levels in the MRA. Specifically, this research different from previous studies is to show the clear comparison of the performance on noise suppression based on two rules of the multi-level wavelet transform (MLWT) and single connected-unit level wavelet transform (SC-ULWT) from the OVT of a PEM fuel cell. For MLWT-based noise suppression, without use of detail component Dn , only approximation component decomposed at the previous level is considered as an input at the next decomposition level. The reconstruction process is vice versa. On the contrary, the SC-ULWT performed the MRA and noise suppression at every unit decomposition and reconstruction levels. As it were, the noise suppressed OVT at the previous unit level is used an input at the next unit level. This rule applied n unit level WT that suppressed sensing of noisy at every steps and connected in series. This research compared the performance on noise suppression between the MLWT and SC-ULWT by the signal-to-noise (SNR) ratio calculation. The comparative analysis on noise suppression between two rules clearly showed in this research. For reference, the PEM fuel cell for this research was made by the ‘Materials and Electro-Chemistry Laboratory’ in Inha University.

WAVELET TRANSFORMATION FOR ENHANCING MAMMOGRAPHIC IMAGES

Mammographic images are often prone to noises and consequently make the task of radiologist to come up with the precise diagnosis. Though there are several denoising techniques for the same is available but while denoising they often suffers from the problem of eliminating the micron level details in the noise influenced images. It’s a trade-off which prohibits efficient micro-classification of mammary tissues. This, in this study we present a solution for the same by utilizing multi level wavelet transformation to enable preservation of micron level details in the images.

A Novel Image Encryption Supported by Compression Using Multilevel Wavelet Transform

In this paper we propose a novel approach for image encryption supported by lossy compression using multilevel wavelet transform. We first decompose the input image using multilevel 2-D wavelet transform, and thresholding is applied on the decomposed structure to get compressed image. Then we carry out encryption by decomposing the compressed image by multi-level 2-D Haar Wavelet Transform at the maximum allowed decomposition level. These results in the decomposition vector C and the corresponding bookkeeping matrix S. The decomposition vector C is reshaped into the size of the input image. The reshaped vector is rearranged by performing permutation to produce encrypted image. The vector C and the matrix S serve as key in the process of both encryption and decryption. In this analysis, we have noticed that the reconstructed image is a close replica of the input image.

Breast volume denoising and noise characterization by 3D wavelet transform

Breast imaging through cone-beam computed tomography provides a digital breast volume, with which the three-dimensional (3D) breast tissues can be analyzed. Data denoising, as a preprocessing step for subsequent volumetric breast segmentation is always needed. In this paper, we report a volumetric denoising technique by a separable 3D wavelet transform (WT), i.e. a ‘2D WT plus 1D WT’ scheme. Specifically, the scheme performs two-dimensional (2D) wavelet denoising on a stack of slice images of the breast volume, followed by one-dimensional (1D) wavelet denoising along the stacking direction. The denoising is achieved by wavelet decomposition, high-pass subband attenuation, and wavelet synthesis. A one-level 3D WT produces eight subbands occupying the octants of the 3D wavelet space. Multilevel WT also provides a multiresolution representation of breast volume, i.e. a sequence of low-pass subbands. In general, most noise and irregularity features are imparted into the high-pass subbands, which are removed or reduced for denoising purpose. Meanwhile, the information in a subband can be characterized in terms of energy, variance, and entropy. Through 3D visualization, the spatial structure in a subband can also be visually perceived. Experimental demonstration with the breast volume reconstructed from a specimen is provided.

Detection of microcalcifications by means of multiscale methods and statistical techniques.

The detection of clustered microcalcifications can help the radiologist to detect early breast cancer. Microcalcifications exhibit some important characteristics, such as small size and high luminosity. Use of a computer-aided diagnosis (CAD) method can prevent them being overlooked. In this report, a multiresolution analysis is performed based on a multilevel wavelet transformation. Decomposition produces sub-band images which become visible only as details of the different scales. Thereafter, all the images will be combined in a final image, in order to obtain an image that contains all the interest details at the scale where microcalcifications tend to appear. Once the image, called detail image, is obtained, it is necessary to determine which details correspond with microcalcifications. Statistical analysis of the histogram permits classification of the zones likely to contain microcalcifications. Applying this statistical techniques over the whole image and representing the results in a two-dimensional map, clustered microcalcification regions are clearly distinguishable.