Search

Graves' disease is a common condition that is diagnosed clinically by determining the smoothness of the thyroid texture and its morphology in ultrasound images. Currently, the most widely used approach for the automated diagnosis of Graves' disease utilizes Convolutional Neural Networks (CNNs) for both feature extraction and classification. However, these methods demonstrate limited efficacy in capturing texture features. Given the high capacity of wavelets in describing texture features, this research integrates learnable wavelet modules utilizing the Lifting Scheme into CNNs and incorporates a parallel wavelet branch into the ResNet18 model to enhance texture feature extraction. Our model can analyze texture features in spatial and frequency domains simultaneously, leading to optimized classification accuracy. We conducted experiments on collected ultrasound datasets and publicly available natural image texture datasets, our proposed network achieved 97.27% accuracy and 95.60% recall on ultrasound datasets, 60.765% accuracy on natural image texture datasets, surpassing the accuracy of ResNet and confirming the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

Photometric visual servoing is considered more precise and more robust than methods that use extracted geometric features, considering that photometric visual servoing uses only pure luminance intensities as indicators of vision. However, this method fails to give the expected results in situations when there are significantly distinct variations in illumination. Generating vision-based control signals through the use of the traditional wavelet approach associated with the lifting scheme can resolve this problem and reduce the considerable processing time required for the creation of interaction matrices. Unlike using traditional Wavelet approaches alone, adding the lifting scheme gives rise to a mathematical method that optimizes calculations and permits real-time wavelet modifications. In this work, a novel idea that exploits the advantages of the traditional wavelet approach associated with the attractive features of the lifting scheme, is proposed and consists of employing wavelet coefficients generated through the lifting approach as input signals in vision-based control. The development of an interaction matrix for the proposed features is presented and simulations of various illumination scenarios are carried out to demonstrate the effectiveness of the proposed technique. [ABSTRACT FROM AUTHOR]

With the development of new technologies, 3D models are becoming increasingly important. They are used to design new models, document cultural heritage and scan valuable artefacts or evidence. They are also used in medicine. For these reasons, they are vulnerable to forgery. Protection against forgery done by encrypting the model or signing it digitally may restrict access to the data or require additional files to store the signatures. A good way to confirm the originality of 3D models is fingerprinting. This technique involves attaching a fragile watermark directly to the watermarked data. In the paper, we propose a new fingerprinting method for 3D models. The method hides the fingerprint in the least significant digits of the coordinates of the selected vertices. The fingerprint is created by calculating the hash-based message authentication code (HMAC) from the model textures and all vertex coordinates except the digits intended to attach the fingerprint. These digits are processed using discrete wavelet transform (DWT). The HMAC is attached to the selected DWT coefficients. The inverse discrete wavelet transform is then performed to obtain the new values of the modified digits. The digits are put back into the 3D model coordinates and the model is reassembled. Verification of the model originality is done according to the used steganographic key and consists of comparing the HMAC value extracted from the fingerprinted model with the HMAC value calculated from it. The same values of both HMAC results indicate that the model has not been modified. The proposed method allows efficient model fingerprinting and detection of changes made to any part of the model. The included fingerprints are transparent – the peak signal-to-noise ratio (PSNR) of a fingerprinted model can reach 150dB and its structural similarity can be over 99.8%.

Securing image transmissions has become critical as the demand for image sharing and storage grows. In some commercial uses, in order to achieve a balance between encryption security and efficiency, some researchers have tried to encrypt only the region of interest of the user in the image. In order to accurately encrypt region of interest images, this paper proposed a region of interest image encryption algorithm based on lifting scheme and object detection. The algorithm automatically identifies the region of interest in the image and encrypts it securely and effectively. Firstly, the region of interest in the image is detected using the object detection model, and the non-maximum suppression algorithm is modified to solve the problem that the detection box outputted by the object detection model does not completely contain the region of interest. Then the existing pixel extraction method for region of interest images is improved to extract the pixels from the region of interest images at a faster speed, which improves the overall efficiency of the algorithm. Finally, based on the thought of wavelet lifting transform, combined with chaotic system, a two-layer hybrid lifting scheme encryption algorithm is proposed to encrypt the extracted pixels. Experimental results and security analysis show that the algorithm proposed in this paper can effectively protect all objects at once with high security. [ABSTRACT FROM AUTHOR]

An integer wavelet transform (IWT) based image steganography technique is proposed to prevail the embedding capacity constraints in this domain. Quotient value differencing (QVD) and 2-least significant bit (2-LSB) techniques are employed in the high frequency subbands. It divides the bit region of each coefficient of high-frequency subbands into two halves and two techniques are used concurrently in each half of the bitplane. The high-frequency subbands are divided into non-overlapping 3 × 3 blocks and confidential messages are concealed in eight directions within each block. The low-frequency subband is fragmented into non-overlapping 2 × 2 blocks. Since the human eye cannot tolerate more changes in low-frequency coefficients, one-fourth of the approximation subband (LL) is used to hide confidential messages. The method attains an average embedding capacity (AEC) of 3.14 bpp. To increase the safety of the technique, the confidential messages are hidden in a non-sequential way. The work resists the steganalysis test of StegExpose and outperforms the tests of the standard tool StirMark benchmark 4.0. [ABSTRACT FROM AUTHOR]

Wavelet transform is an important tool in the field of multimedia signal processing. Due to the growing demand for hardware-based design in the field of image, video, and speech signal processing, several researchers have implemented different discrete wavelet transform (DWT) architectures and measured their performances. In this paper, FPGA implementation of compact and low-power VLSI architectures for one-dimensional and two-dimensional DWT and inverse discrete wavelet transform (IDWT) based on lifting and lattice schemes have been proposed. The adder–subtractor-based design approach has been used to replace the multiplier without affecting the precision. Hardware and timing analysis of existing and proposed 1D and 2D architectures are presented here. The proposed and existing DWT and IDWT architectures are also simulated and synthesized in a Virtex-4 FPGA device. The number of slice LUTs, slice registers, clock frequency, delay, power, and power-LUTs product (PLP) of proposed DWT and IDWT architectures based on lifting and lattice schemes have been found using Xilinx Vivado synthesis tool. The proposed design is also implemented on Basys 3 Artix-7 FPGA board and synthesis results show a significant improvement in terms of area and power compared to other existing architectures. [ABSTRACT FROM AUTHOR]

The Cahn-Allen equation is a reaction-diffusion equation of mathematical physics that describes the process of phase separation in multi-component alloy systems and orderdisorder transitions. This paper presented a numerical solution of Cahn-Allen equations by Lifting scheme using different wavelet filter coefficients. The numerical results obtained using this scheme are compared with the exact solution to demonstrate the accuracy and fast convergence in less computational time than the existing scheme. Some problems are taken to show the validity and applicability of the scheme. [ABSTRACT FROM AUTHOR]

Wavelet-based graph neural networks have received increasing attention in the node classification task. Existing graph wavelet-based approaches, however, are not applicable to arbitrary graphs as they use predefined wavelet filters with built-in homophilic assumptions and disregard heterophily. Recent studies attempted to address this issue through a wavelet lifting transform, which requires a bipartite graph, therefore altering the graph topology and leading to undesirable wavelet filters. This paper proposes a generalized adaptive graph wavelet network that preserves the graph topology through computational trees while implementing the lifting scheme on arbitrary graphs. Moreover, this locally defined lifting scheme integrates both high-pass and low-pass frequency components to further enhance feature representation. Finally, we benchmark our model using nine homophilic and heterophilic datasets, and the results demonstrate the effectiveness of our method. • We propose a novel framework for lifting-based adaptive wavelet networks. • Our framework uses local lifting scheme on computational trees, preserving structure. • The fusion operation combines levels of information, enhancing node representation. • Our experiments on nine datasets validate the effectiveness of our method. [ABSTRACT FROM AUTHOR]

An Inertial Navigation System (INS) is a standalone system that provides continuous navigation information on the position, velocity, and attitude in various conditions. Gyroscopes and accelerometers are essential components of INS. However, these sensors are prone to errors that accumulate over time and degrade the accuracy of the INS. Therefore, enhancing the performance of INS is a challenging problem. This paper aims to address this problem by proposing a method to reduce errors in real-time processing. The method uses Lifting Wavelet Transform (LWT), a popular denoising technique that belongs to the wavelet family and has an advantage over the classic wavelets in terms of time and computation complexity. Moreover, the method employs an optimization technique based on the Genetic algorithm as an intelligent averaging for combining different multi-level LWTs, which significantly improves the accuracy and efficiency of INSs. Finally, a new and implementable structure is introduced for denoising inertial sensors. The results show that the proposed method achieves an 83% and 59% improvement in gyroscope and accelerometer data for dynamic data, compared with raw data. Besides, for static data, there are 71% and 36% improvements in gyroscope and accelerometer data, respectively.

An Inertial Navigation System (INS) is a standalone system that provides continuous navigation information on the position, velocity, and attitude in various conditions. Gyroscopes and accelerometers are essential components of INS. However, these sensors are prone to errors that accumulate over time and degrade the accuracy of the INS. Therefore, enhancing the performance of INS is a challenging problem. This paper aims to address this problem by proposing a method to reduce errors in real-time processing. The method uses Lifting Wavelet Transform (LWT), a popular denoising technique that belongs to the wavelet family and has an advantage over the classic wavelets in terms of time and computation complexity. Moreover, the method employs an optimization technique based on the Genetic algorithm as an intelligent averaging for combining different multi-level LWTs, which significantly improves the accuracy and efficiency of INSs. Finally, a new and implementable structure is introduced for denoising inertial sensors. The results show that the proposed method achieves an 83% and 59% improvement in gyroscope and accelerometer data for dynamic data, compared with raw data. Besides, for static data, there are 71% and 36% improvements in gyroscope and accelerometer data, respectively. [ABSTRACT FROM AUTHOR]

Recently Wavelet theory is an advanced tool in science and engineering research; mainly wavelets are applicable in fast algorithms for easy implementation. In this paper, we present an innovative wavelet technique that is lifting scheme for the numerical solutions of hyperbolic differential equations come across in mathematical physics. Here we used orthogonal and biorthogonal wavelets to expose the efficiency and effectiveness of the proposed scheme. The proposed scheme speeds up the convergence in less computational time, which is illustrated through the numerical examples. [ABSTRACT FROM AUTHOR]

Copyright of Przegląd Elektrotechniczny is the property of Przeglad Elektrotechniczny and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Discrete Wavelet Transform (DWT) is considered among the few computationally expensive block of multimedia compression standards. In this work, we proposed a reduced hardware complexity VLSI architectures of 5/3 and 9/7 lifting bi-orthogonal DWT for multimedia applications. The architecture uses a combination of Distribute Arithmetic (DA) and Canonical Signed Digit (CSD) based implementation to reduce the hardware complexity. The resulting lifting based architecture discretely finds optimized number of sum of products to give minimum realization. The architecture is implemented on Field Programmable Gate Array (FPGA) with results compared with known classical and other optimized DWT architectures. • A reduced hardware complexity architecture of 5/3 and 9/7 lifting bi-orthogonal DWT • Uses Distribute Arithmetic and Canonical Signed Digit to give minimum realization • Functionality verified using Altera DE2-115 Cyclon IV FPGA • Hardware results compared with other optimized DWT architectures [ABSTRACT FROM AUTHOR]

In the secured data communication scenario, image based Reversible Data Hiding (RDH) at transform domain is receiving its due popularity now. In research works so far, evaluation of transforms in RDH is very minor compared to that of image coding. Since both the areas deal primarily with integers, the present study takes an appreciable step forward by investigating the utilization of integer-to-integer wavelet transforms that have reversibility, which helped in developing high capacity and robust RDH algorithms. The proposed investigation tends to bridge gaps in the area of data hiding with the available coding transforms. In this work, integer wavelet transforms like Haar, 5/3, 2/6, 9/7-M, 2/10, 5/11-C, 5/11-A, 6/14, 13/7-T, 13/7-C and 9/7-F are evaluated using a generalized threshold-based histogram shifting technique. This work compares the performance of all integer wavelet transforms and other state of the art techniques with respect to their embedding capacity and image visual quality. This analysis leads to a better understanding of the relationship between the embedding capacity and the stego image quality whenever different wavelets were utilized.

This paper interprets the application of second-generation wavelet transform (SGWT) to IEEE-14 test bus system combined with renewable source. The goodness of this lifting-scheme-driven second-generation wavelet transform is that it is a time domain analysis which needs less processing time and thus less memory. On the other hand, the SGWT is very simple and easy to implement. The IEEE-14 bus system is designed with the photovoltaic system from which different distorted signals are captured. The PV disconnection is carried out within the power quality (PQ) environment. Further, this paper discriminates the PV disconnection events from PQ events by selecting threshold value. For this negative sequence, voltage signal has been extracted at point of common coupling. The MATLAB simulation results show that SGWT has provided good results. Moreover, SGWT is also found more robust. Moreover, data mining-based approach, named as random forest (RF), has also implemented for discrimination of PV disconnection events from PQ disturbances. [ABSTRACT FROM AUTHOR]

Color filter array is a spatial multiplexing of pixel-sized filters fabricated over pixel sensors in most color image sensors. The state-of-the-art lossless coding techniques of raw sensor data captured by such sensors leverage spatial or cross-color correlation using lifting schemes. In this paper, we propose a lifting-based lossless white balance algorithm. When applied to the raw sensor data, the spatial bandwidth of the implied chrominance signals decreases. We propose to use this white balance as a pre-processing step to lossless CFA subsampled image/video compression, improving the overall coding efficiency of the raw sensor data. [ABSTRACT FROM AUTHOR]

Lifting-based wavelet transform has been extensively used for efficient compression of various types of visual data. Generally, the performance of such coding schemes strongly depends on the lifting operators used, namely the prediction and update filters. Unlike conventional schemes based on linear filters, we propose, in this paper, to learn these operators by exploiting neural networks. More precisely, a classical Fully Connected Neural Network (FCNN) architecture is firstly employed to perform the prediction and update. Then, we propose to improve this FCNN-based Lifting Scheme (LS) in order to better take into account the input image to be encoded. Thus, a novel dynamical FCNN model is developed, making the learning process adaptive to the input image contents for which two adaptive learning techniques are proposed. While the first one resorts to an iterative algorithm where the computation of two kinds of variables is performed in an alternating manner, the second learning method aims to learn the model parameters directly through a reformulation of the loss function. Experimental results carried out on various test images show the benefits of the proposed approaches in the context of lossy and lossless image compression. [ABSTRACT FROM AUTHOR]

Computed Tomography (CT) is a medical imaging technique that is being prominently used in the healthcare domain to obtain a detailed view of the body for diagnostic purposes of various diseases. This form of medical imaging is associated with high ionizing radiations that are powerful enough to penetrate through the body to create images on the computer screen. But, multiple exposures to such high dose ionizing radiation could raise the chances of cancer and could be fatal for patients diagnosed with cancer. However, to reduce the radiation dosage associated with CT images, low-dose CT (LDCT) is being used for medical screening in recent days. The United States Preventive Services Task Force guidelines has revealed that LDCT screening reduces the mortality rate of patients and can be considered safe comparatively. But, as radiation dosage is reduced, LDCT images are corrupted with noise and artifacts which affect the visibility of the medical image. This in turn can affect the decision of the radiologists. Therefore, LDCT images are required to be de-noised before being used for diagnosis to improve the image quality and elevate visibility. This work is one such method where a combination of modified Discrete Wavelet Transform (DWT) and Goodness of Fit shrinkage (GoFShrink) thresholding technique is used to denoise the LDCT images to enhance the quality for diagnostic purposes. The modified DWT is associated with a lifting scheme that provides memory for in-place arithmetic operations, flexible factorization of the 2-channel filter banks as well as robustness with exact reversible reconstruction and enhances the performance of the wavelet transform (WT). The PSNR, MSE, SSIM, and SNR values of the de-noised images are calculated and a comparative analysis is performed with the conventional de-noising techniques. A comparative analysis between the various shrinkage techniques is also performed. The simulation results show an increase in PSNR and SNR values when compared with the conventional methods. [ABSTRACT FROM AUTHOR]

The need to detect and classify objects correctly is a constant challenge, being able to recognize them at different scales and scenarios, sometimes cropped or badly lit is not an easy task. Convolutional neural networks (CNN) have become a widely applied technique since they are completely trainable and suitable to extract features. However, the growing number of convolutional neural networks applications constantly pushes their accuracy improvement. Initially, those improvements involved the use of large datasets, augmentation techniques, and complex algorithms. These methods may have a high computational cost. Nevertheless, feature extraction is known to be the heart of the problem. As a result, other approaches combine different technologies to extract better features to improve the accuracy without the need of more powerful hardware resources. In this paper, we propose a hybrid pooling method that incorporates multiresolution analysis within the CNN layers to reduce the feature map size without losing details. To prevent relevant information from losing during the downsampling process an existing pooling method is combined with wavelet transform technique, keeping those details "alive" and enriching other stages of the CNN. Achieving better quality characteristics improves CNN accuracy. To validate this study, ten pooling methods, including the proposed model, are tested using four benchmark datasets. The results are compared with four of the evaluated methods, which are also considered as the state-of-the-art. [ABSTRACT FROM AUTHOR]

This paper presents a new multiscale method for analysing water pollutant data located in river networks. The main idea of the proposed method is to adapt the conventional lifting scheme, reflecting the characteristics of streamflow data in the river network domain. Due to the complexity of the data domain structure, it is difficult to apply the lifting scheme to the streamflow data directly. To solve this problem, we propose a new lifting scheme algorithm for streamflow data that incorporates flow‐adaptive neighbourhood selection, flow proportional weight generation and flow‐length adaptive removal point selection. A nondecimated version of the proposed lifting scheme is also provided. The simulation study demonstrates that the proposed method successfully performs a multiscale analysis of streamflow data. Furthermore, we provide a real data analysis of water pollutant data observed on the Geum‐River basin compared to the existing smoothing method. [ABSTRACT FROM AUTHOR]

Image compression is the most important image processing method extensively deployed in different appliances. "Discrete wavelet transform (DWT)" is one of the well‐adopted transforming methods exploited for compressing images. The extremely deployed version of DWT is convolution‐oriented. Nevertheless, the lifting‐oriented DWT scheme requires more contemplation on more proficient performance and lesser computation cost. This paper intends to propose a deep learning‐based image compression model with a lifting scheme for predicting high‐frequency subbands. Moreover, the fine‐tuning in lifting factorization is done by a new Sea Lion with Averaged Update Evaluation that includes new cosine estimation under the COordinate Rotation DIgital Computer algorithm. Similarly, this study defines a new single objective function that merges the multiconstraints, like, "Peak Signal to Noise Ratio (PSNR) and Compression Ratio (CR)". At last, the supremacy of the presented approach is proved with respect to varied measures, like, CR, PSNR and so on. [ABSTRACT FROM AUTHOR]

Wavelet theory is a recently developed promising tool in science and engineering research; particularly wavelets are successfully used in fast algorithms for easy implementation. In this paper, we present wavelet-based numerical solutions of elliptic boundary value problems encountered in mathematical physics, using lifting scheme. Here we used orthogonal and biorthogonal wavelets to demonstrate the effectiveness and efficiency of the proposed technique. The proposed scheme speeds up convergence in lesser computational time. Numerical examples are given to prove the robustness of proposed technique. [ABSTRACT FROM AUTHOR]

Recently, wavelet theory has become a well recognized promising tool in science and engineering field; especially, wavelets are successfully used in fast algorithms for easy execution. In this paper, we developed wavelet lifting scheme using orthogonal and biorthogonal wavelets for the numerical solution of dynamic Reynolds equation for micropolar fluid lubrication. The numerical results gained through proposed scheme are compared with the exact solution to expose the accuracy with speed of convergence in lesser computational time as compared with the existing methods. The examples are given to demonstrate the applicability and attractiveness of proposed method. [ABSTRACT FROM AUTHOR]

The on-board object detection of remote sensing images is of great significance and challenges in both military and civilian fields. To achieved on-board object detection, we require a high-performance and low power hardware architecture. Inspired by the hardware implementation advantages of the lifting scheme in JPEG2000, we introduce the lifting scheme into object detection networks to promote the hardware implementation capability. Based on the equivalence between the lifting scheme and the convolution, we proposed a lifting based FPGA accelerator. The lifting scheme provides a parallel strategy for every convolutional layer to obtain high performance, and high resource utilization. The proposed accelerator can improve the throughput of networks, and promote the utilization efficiency of hardware resources. The experimental results on lifting based YOLOv3-tiny (LS-YOLOv3-tiny) show that the proposed accelerator has better power efficiency than other hardware implementation, and have better DSP efficiency than other FPGA-based designs.

In the secured data communication scenario, image based Reversible Data Hiding (RDH) at transform domain is receiving its due popularity now. In research works so far, evaluation of transforms in RDH is very minor compared to that of image coding. Since both the areas deal primarily with integers, the present study takes an appreciable step forward by investigating the utilization of integer-to-integer wavelet transforms that have reversibility, which helped in developing high capacity and robust RDH algorithms. The proposed investigation tends to bridge gaps in the area of data hiding with the available coding transforms. In this work, integer wavelet transforms like Haar, 5/3, 2/6, 9/7-M, 2/10, 5/11-C, 5/11-A, 6/14, 13/7-T, 13/7-C and 9/7-F are evaluated using a generalized threshold-based histogram shifting technique. This work compares the performance of all integer wavelet transforms and other state of the art techniques with respect to their embedding capacity and image visual quality. This analysis leads to a better understanding of the relationship between the embedding capacity and the stego image quality whenever different wavelets were utilized. [ABSTRACT FROM AUTHOR]

Image watermarking is a technique to embed a piece of information or data, which also can be an image, into the host or medium image for specific purposes, such as authentication, security, or copyright protection. A good image watermarking method should have a good perceptibility after adding the watermark. It should also robust against attacks and the watermark should be able to be recovered although the watermarked image has distorted. This paper presents the proposed lifting image watermarking scheme based on wavelet transform through singular value and Schur decomposition. This method is proposed to improve the perceptibility and the robustness of the image watermarking scheme, especially against non-geometrical attacks. The lifting scheme is performed by lifting wavelet transform to produce a better imperceptibility. The use of singular value decomposition is used to increase the intactness of the embedded watermark from the distorted image, which is performed on the triangular matrix of lowfrequency sub-band produced by Schur decomposition. In perceptibility performance, experimental results show that the proposed method could produce a watermarked image very close to the host image with more than 99% of similarity level. In robustness performance, the proposed method can recover the embedded watermark even though the watermarked image has been distorted by non-geometrical attacks with more than 90% of normalized correlation average. [ABSTRACT FROM AUTHOR]

• Directional selectivity of discrete wavelet transform significantly improved. • Nearly isotropic signal decomposition into 12 directions. • More efficient trade-off between computational cost and directional selectivity. • Designed in fully discrete setting and is easy to implement in spatial domain. • Both global and local edge structures are successfully detected. In this paper, we propose a new two-dimensional directional discrete wavelet transform that can decompose an image into 12 multiscale directional edge components. The proposed transform is designed in a fully discrete setting and thus is easy to implement in actual computations. The proposed transform is viewed as a category of redundant discrete wavelet transforms implemented by fast in-place computational algorithms by a lifting scheme that has been modified to incorporate redundancy. The redundancy is limited to (N × J + 1) / 4 , where N = 12 is the directional selectivity and J is a decomposition level of the multiscale transform. Numerical experiments in edge analysis using various images demonstrate the advantages of the proposed method over some conventional standard methods. The proposed method outperforms several conventional edge detection methods in identifying both the location and orientation of edges, and well captures the directional and geometrical features of images. [ABSTRACT FROM AUTHOR]

A new and an enriched JPEG algorithm is provided for identifying redundancies in a sequence of irregular noisy data points which also accommodates a reference-free criterion function. Our main contribution is by formulating analytically (instead of approximating) the inverse of the transpose of JPEG-wavelet transform without involving matrices which are computationally cumbersome. The algorithm is suitable for the widely-spread situations where the original data distribution is unobservable such as in cases where there is deficient representation of the entire population in the training data (in machine learning) and thus the covariate shift assumption is violated. The proposed estimator corrects for both biases, the one generated by endogenous truncation and the one generated by endogenous covariates. Results from utilizing 2 000 000 different distribution functions verify the applicability and high accuracy of our procedure to cases in which the disturbances are neither jointly nor marginally normally distributed.