The proceedings contain 56 papers. The topics discussed include: review of recent results on optimal orthonormal subband coders;comparison of waveletimage coding schemes for seismic data compression;image quality mea...
The proceedings contain 56 papers. The topics discussed include: review of recent results on optimal orthonormal subband coders;comparison of waveletimage coding schemes for seismic data compression;image quality measurement using the Haar wavelet;lossless image compression using wavelets over finite rings and related architectures;on consistent signal reconstruction from wavelet extrema representation;seismic imaging in wavelet domain: decomposition and compression of imaging operator;application of differential mapping and wavelet transform;usage of short wavelets for scalable audio coding;enhanced resolution control for video sequences;regularized multiresolution methods for astronomical image enhancement;weighted time-frequency and time-scale transforms for non-stationary signal detection;and a wavelet detector for distributed objects.
Due to the irrational coefficients, the orthogonal wavelet filter banks (FBs) need a lot of resources when implemented on hardware. As a result, there is a decrease in operating speed, a significant memory requirement...
详细信息
Due to the irrational coefficients, the orthogonal wavelet filter banks (FBs) need a lot of resources when implemented on hardware. As a result, there is a decrease in operating speed, a significant memory requirement, and an increase in power consumption. This paper proposes an orthogonal low complex symmetric Daubechies-2 (4-tap) wavelet filter bank (FB) to address these problems. This is accomplished by obtaining dyadic filter coefficients and making the suggested FB symmetric by marginally modifying the perfect reconstruction (PR) criterion. By using fewer adders and shifters, the suggested wavelet FB achieves a significant reduction in dynamic power consumption without the need for multipliers. This is confirmed by implementing the suggested wavelet FB on the Zedboard ZYNQ-7000 AP-SoC (Zynq FPGA from Xilinx) field programmable gate array (FPGA). The suitability of the suggested FB is tested in medical image retrieval and image compression applications. Results from simulations demonstrate that, when compared to state-of-the-art techniques, the suggested wavelet FB performs better in terms of retrieval accuracy (ARP, ARR) for medical image retrieval and PSNR for image compression on the benchmark image datasets.
This paper begins by examining the definition and fundamental properties of the two-dimensional linear canonical wavelet transform (2-D LCWT) within the framework of linear canonical transform theory. The LCWT offers ...
详细信息
This paper begins by examining the definition and fundamental properties of the two-dimensional linear canonical wavelet transform (2-D LCWT) within the framework of linear canonical transform theory. The LCWT offers significant advantages in handling multi-scale and multi-directional signals. Building on this, a novel model is proposed based on the parametric Riesz transform and parametric monogenic signal, introducing a parametric monogenic linear canonical wavelet along with its associated transform, referred to as PMLW. Through parametric embedding within the monogenic linear canonical wavelet framework, the proposed transform achieves enhanced flexibility and robustness, facilitating more efficient analysis of intricate features in complex signals. This approach leverages 2-D analytical signal theory to incorporate phase information with directional characteristics, thereby preserving rich directional details in multi-scale analysis. Furthermore, the potential applications of PMLW in image denoising tasks are explored, demonstrating its effectiveness in preserving structural details while suppressing noise.
In addressing temporal dependencies within data, specifically in signal analysis, the integration of Deep Neural Networks (DNN) has demonstrated notable improvements when coupled with a preprocessing stage designed fo...
详细信息
In addressing temporal dependencies within data, specifically in signal analysis, the integration of Deep Neural Networks (DNN) has demonstrated notable improvements when coupled with a preprocessing stage designed for extracting implicit information. In this context, the widely adopted wavelet Transform (WT) has garnered attention for its remarkable results. However, inherent challenges, such as the imperative definition of parameters for optimal information extraction across diverse scales and resolutions, as well as the prerequisite batch conversion of signals prior to network training, underscore the need for innovative solutions. In response to these challenges, the main contribution of this manuscript is a novel DNN architecture to replace the preprocessing phase. This architecture produces output characteristics resembling those derived from WT, preventing the necessity for a preceding batch execution. Our contribution not only stands as an independent solution but also seamlessly integrates with other modeling techniques, eliminating the prerequisite for the upfront execution of any wavelet transformations. To assess its performance, our methodology undergoes rigorous evaluation against DNNs in classifying signals from real-world applications. Our findings indicate the promising potential of end-to- end schemes in advancing signal analysis applications.
In this article, a simplified efficient 2-D discrete wavelet transform (DWT) architecture based on the lifting scheme is presented. By approximating the multipliers required for the multiplication operations in the Co...
详细信息
In this article, a simplified efficient 2-D discrete wavelet transform (DWT) architecture based on the lifting scheme is presented. By approximating the multipliers required for the multiplication operations in the Cohen-Daubechies-Feauveau (CDF) 9/7 filter, computational resources are significantly reduced. In addition, the strip-based scanning method requires minimal storage resources;for single-level 2-D DWT, there is no need for RAM resources that are dependent on image size. This characteristic makes the proposed architecture particularly well-suited for applications in wireless visual sensor networks (WvSNs). The proposed 2-D DWT and IDWT hardware implementations ensure a reconstructed quality exceeding 90.56-dB peak signal-to-noise ratio (PSNR) for various N x N images with low power consumption. Compared to existing 2-D DWT architectures, our design offers significant advantages, particularly when processing larger block sizes. The area-delay product (ADP) is reduced by at least 10.02%, the energy per image (EPI) decreases by 75.98%, and power consumption is lowered by 53.52%. Furthermore, the circuit performance in the multilevel architecture is outstanding, and the savings in required resources and energy make the proposed architecture highly suitable for applications in WvSNs.
This study focuses on classifying nine different types of fruit using several deep learning architectures, including AlexNet, ResNet-50, GoogleNet, DenseNet-201, and EfficientNet-b0, while also investigating the effec...
详细信息
This study focuses on classifying nine different types of fruit using several deep learning architectures, including AlexNet, ResNet-50, GoogleNet, DenseNet-201, and EfficientNet-b0, while also investigating the effect of wavelet Transform (WT) on the classification success rates of these models. The research begins by testing these architectures on a clean, well-curated dataset, where AlexNet, ResNet, GoogleNet, DenseNet, and EfficientNet achieve impressive success rates of 93.88%, 100%, 99.34%, 99.94%, and 99.88%, respectively. However, when the models are tested on an online dataset of fruit images sourced from the internet to reflect better real-life conditions, the success rates drop significantly to 63.10%, 76.35%,74.64%, 76.63%, and 73.07%, respectively. This drop highlights the challenges posed by real-world noise, varying image backgrounds, and inconsistencies in data quality. To address this issue, the study incorporates WT as a preprocessing step and finds that it significantly improves classification accuracy, particularly for the GoogleNet architecture. The results emphasize the importance of considering real-world complexities such as background noise and image distortion when developing classification models. Furthermore, the study demonstrates the potential of WT to increase classification performance, especially in specific architectures such as GoogleNet, which shows the most significant improvement. In addition, the effect of WT was statistically analyzed and the results were evaluated. Overall, this research contributes to a deeper understanding of the factors that affect the performance of deep learning models in fruit classification tasks and provides valuable insights into optimizing these models for more robust real-world applications.
Approximate circuits play a vital role in enhancing efficiency and optimizing resource use in modern computing systems. Their benefits are particularly notable in fields that tolerate minor inaccuracies, such as image...
详细信息
Approximate circuits play a vital role in enhancing efficiency and optimizing resource use in modern computing systems. Their benefits are particularly notable in fields that tolerate minor inaccuracies, such as imageprocessing, signalprocessing, and data mining, where a slight reduction in precision can lead to substantial savings in power and space requirements. This study explores an innovative design for an approximate full subtractor based on the principle of pruning, meticulously implemented using universal two-input NOR gates, valued for their cost efficiency, low power consumption, and compact design. Existing approximate subtractors have been designed using non-universal basic gates such as XOR, XNOR, NOT, and AND gates. In contrast, the proposed approach utilizes only the universal NOR gate, leading to improved circuit efficiency in terms of area, delay, and power consumption. Additionally, this work evaluates performance metrics of approximate circuits, demonstrating their effectiveness in various imageprocessingapplications involving full subtractors.
Graphic design emphasizes constructing visual aesthetics using computer-aided platforms, effectively shaping various design elements. A wavelet-based imageprocessing technology solution is proposed for graphic design...
详细信息
Deep learning-based acoustic echo cancellation (AEC) systems have advanced significantly, yet previous methods often rely on a single transform, such as short-time fourier transform (STFT) or constant Q transform, whi...
详细信息
Deep learning-based acoustic echo cancellation (AEC) systems have advanced significantly, yet previous methods often rely on a single transform, such as short-time fourier transform (STFT) or constant Q transform, which limits feature richness and leads to heavy models. In contrast, this paper introduces a novel feature-fusion-based encoder-decoder with a sparse masked attention network, specifically designed to enhance echo and background noise suppression. Our model uniquely combines discrete wavelet transform (DWT) and STFT features, leveraging both transforms to achieve richer feature representation. By employing the Daubechies wavelet ("db4"), the model attains high time resolution for high frequencies and improved frequency resolution for low frequencies-crucial for effective noise cancellation. The STFT component captures temporal spectral content, complementing DWT's strengths. To handle double-talk scenarios, a sparse masked attention mechanism selectively focuses on relevant signal windows, reducing computational load while enhancing accuracy. This masked network enables a causal model suitable for real-time applications. Additionally, Smooth L1 loss promotes stable convergence during training. Experimental results on the AEC challenge dataset demonstrate that our model outperforms traditional methods, achieving superior echo return loss enhancement, perceptual evaluation of speech quality, and correlation coefficient, validating its effectiveness in robust echo cancellation and speech quality improvement.
Digital watermarking plays a vital role in safeguarding data integrity and intellectual property, particularly in medical imaging, where ensuring authenticity and preserving privacy are paramount. Embedding hidden inf...
详细信息
Digital watermarking plays a vital role in safeguarding data integrity and intellectual property, particularly in medical imaging, where ensuring authenticity and preserving privacy are paramount. Embedding hidden information within images aids authentication, copyright protection, and integrity verification. This work proposes a robust palmprint-based watermarking approach based on adaptive ACM watermarking with anon- linear equation technique to enhance patient security. The proposed method integrates local binary patterns and histograms of oriented gradients for robust feature extraction, discrete wavelet transforms for multilevel image decomposition, Arnold Cat Map for chaotic mapping, and singular value decomposition for stable watermark embedding. The contribution goals are improving security, imperceptibility, and resilience against imageprocessing attacks. Experimental results demonstrate the method's effectiveness, with a peak signal-tonoise ratio of 63.52 and a structural similarity index of 1.00, indicating high image quality retention. An equal error rate of 0.035 confirms the method's reliability in watermark detection. These findings underscore the proposed method's suitability for secure medical image watermarking applications, enhancing data integrity and patient privacy.
暂无评论