Digital Signal Processing最新文献_第4页

TMN: Transformer in matrix network for single image super-resolution with enhanced shallow feature preservation

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-31 DOI: 10.1016/j.dsp.2025.105207

Ou Ao , Zhenhong Shang

{"title":"TMN: Transformer in matrix network for single image super-resolution with enhanced shallow feature preservation","authors":"Ou Ao , Zhenhong Shang","doi":"10.1016/j.dsp.2025.105207","DOIUrl":"10.1016/j.dsp.2025.105207","url":null,"abstract":"<div><div>Transformer-based image super-resolution has witnessed remarkable advancements in recent years. However, as transformer networks grow in depth, numerous existing super-resolution methods encounter challenges in effectively preserving shallow features, which play a crucial role in single image super-resolution. The low-resolution input image contains crucial structural and contextual information, and the shallow features serve as the carriers of this information. To address the challenge of preserving shallow features, we propose the Transformer in Matrix Network (TMN), a novel architecture specifically tailored for single image super-resolution. TMN incorporates a redesigned and optimized matrix mapping module, which arranges transformer blocks in a matrix structure to preserve and effectively exploit shallow features while facilitating the efficient reuse of hierarchical feature representations across the network. Additionally, TMN refines the efficient transformer to augment its capacity for modelling long-range dependencies, thereby enabling enhanced integration of information from spatially correlated regions within the image. To further enhance the reconstruction performance, TMN incorporates the structural loss into the loss function. By constraining the relevant statistical quantities, it improves the perceptual fidelity and preserves the intricate details. Experimental results show that TMN achieves competitive performance, with a reduction in computational costs by approximately one-third compared to leading methods like SwinIR. TMN's efficient design and high-quality reconstruction make it particularly suitable for deployment on resource-constrained devices, addressing a critical need in practical applications. The implementation code is publicly available at <span><span>https://github.com/13752849314/TMN</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"162 ","pages":"Article 105207"},"PeriodicalIF":2.9,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143746338","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Performance analysis of active STAR-RIS-assisted NOMA with hardware impairments at finite blocklength

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-31 DOI: 10.1016/j.dsp.2025.105206

Shiv Kumar, Brijesh Kumbhani

{"title":"Performance analysis of active STAR-RIS-assisted NOMA with hardware impairments at finite blocklength","authors":"Shiv Kumar, Brijesh Kumbhani","doi":"10.1016/j.dsp.2025.105206","DOIUrl":"10.1016/j.dsp.2025.105206","url":null,"abstract":"<div><div>The active simultaneously transmitting and reflecting-reconfigurable intelligent surface's (ASTAR-RIS) potential to avoid multiplicative fading loss by utilizing integrated reflection-type amplifiers has drawn considerable interest. This paper investigates the finite blocklength (FBL) analysis of ASTAR-RIS-assisted non-orthogonal multiple access (NOMA) with perfect and imperfect successive interference cancellation (SIC) in the presence of hardware impairments over cascaded Rician fading channels. Firstly, we derive the statistical distribution of cascaded Rician fading channels with the help of Laguerre polynomial series approximation. Secondly, we derive the novel analytical expression for average block error rate (ABLER), ergodic rate (ER), and system throughput with the help of the Gauss Chebyshev quadrature and the Gauss Laguerre quadrature method. Thirdly, the asymptotic expression for ABLER is derived to gain useful insights. Finally, the Monte Carlo simulations are used to verify the analytical results. Numerical results verify the correctness and superior performance of ASTAR-RIS-assisted NOMA (ASTAR-RIS-NOMA) over passive STAR-RIS-assisted NOMA (PSTAR-RIS-NOMA) and ASTAR-RIS-assisted orthogonal multiple access (OMA) (ASTAR-RIS-OMA). Additionally, the impact of other system parameters like imperfect SIC, hardware impairments, and block length are analyzed.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"163 ","pages":"Article 105206"},"PeriodicalIF":2.9,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143759201","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Dual-stream feature pyramid network with task interaction for underwater object detection

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-31 DOI: 10.1016/j.dsp.2025.105199

Wenming Zhang , Haobo Wang , Haibin Li , Yaqian Li , Tao Song , Haoran Liu , GuanCheng Wang

{"title":"Dual-stream feature pyramid network with task interaction for underwater object detection","authors":"Wenming Zhang , Haobo Wang , Haibin Li , Yaqian Li , Tao Song , Haoran Liu , GuanCheng Wang","doi":"10.1016/j.dsp.2025.105199","DOIUrl":"10.1016/j.dsp.2025.105199","url":null,"abstract":"<div><div>Underwater object detection presents significant challenges due to factors such as image degradation caused by light absorption and scattering, insufficient multi-scale feature representation, and the high similarity between objects and the background. This paper proposes a novel underwater object detection method that integrates the Dual-Stream Feature Pyramid Network (DS-FPN) with a Task Interaction Module (TIM) to enhance detection performance in complex underwater environments. The DS-FPN effectively captures multi-scale features through a parallel path fusion strategy, significantly improving both semantic feature extraction and fine-grained object representation. The TIM module dynamically adjusts feature weights by facilitating feature interaction between tasks, allowing the classification and regression tasks to work in tandem and mitigating the trade-off between classification accuracy and object localization precision. Extensive experiments on multiple public datasets demonstrate the effectiveness of the proposed method, which achieves improvements of 2.2%, 2.5%, and 3.5% in mAP on the DUO, UTDAC, and TrashCan datasets, respectively, outperforming existing state-of-the-art methods. Moreover, applying the DS-FPN and TIM modules to various detection frameworks further validates the method's strong generalization capability and its applicability in multi-task scenarios. These results confirm that the proposed approach provides an efficient and versatile solution for advancing underwater object detection performance.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"163 ","pages":"Article 105199"},"PeriodicalIF":2.9,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143759200","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Desmoke-VCU: Improved unpaired image-to-image translation for removing smoke from laparoscopic images

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-28 DOI: 10.1016/j.dsp.2025.105177

Wenjie Wang , Qi Yuan , Pengtai Huang , Xiaohua Wang , Huajian Song

{"title":"Desmoke-VCU: Improved unpaired image-to-image translation for removing smoke from laparoscopic images","authors":"Wenjie Wang , Qi Yuan , Pengtai Huang , Xiaohua Wang , Huajian Song","doi":"10.1016/j.dsp.2025.105177","DOIUrl":"10.1016/j.dsp.2025.105177","url":null,"abstract":"<div><div>In laparoscopic surgery, maintaining a clear field of view is crucial; however, smoke generated during the procedure can impair surgical judgment. Cycle-consistent generative adversarial network (CycleGAN) has been widely applied to image dehazing tasks as it does not require paired smoke and clear images for training. However, it has limited intrinsic smoke removal effectiveness. To improve the smoke removal performance, this paper proposes a smoke removal model named the Desmoke-VCU based on the DeSmoke-LAP. This method combines the UNet and vision transformer (ViT) as generators within the CycleGAN framework and employs self-supervised pre-training techniques to pre-train the generator, aiming to generate realistic smoke-free images. Furthermore, a structural similarity index metric (SSIM) loss function is introduced to preserve textures and details in the generated images. Experimental results demonstrate that the proposed approach enhances the quality and realism of generated images, maintains strong correlations between smokey and smoke-free images, and exhibits superiority over recent models proposed in this field.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"162 ","pages":"Article 105177"},"PeriodicalIF":2.9,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Orbital angular momentum radar detection characteristics analysis based on ambiguity function

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-28 DOI: 10.1016/j.dsp.2025.105197

Yin Dang, Deyu Li, Yusheng Fu, Yi Liao

引用次数: 0

Minimum error entropy with affine projection algorithm for robust adaptive filtering

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-28 DOI: 10.1016/j.dsp.2025.105198

Huaiyuan Zhang , Guoliang Li , Yunxian Hou , Hongbin Zhang , Shan Zhong

引用次数: 0

RIE-GAN: A retrievable image encryption method based on GAN

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-28 DOI: 10.1016/j.dsp.2025.105202

Yang Nan, Yan Wo

{"title":"RIE-GAN: A retrievable image encryption method based on GAN","authors":"Yang Nan, Yan Wo","doi":"10.1016/j.dsp.2025.105202","DOIUrl":"10.1016/j.dsp.2025.105202","url":null,"abstract":"<div><div>With the development of cloud computing, people tend to encrypt images and upload them to the cloud for saving storage space and protecting privacy. However, these image encryption methods will hinder the availability of images such as similarity retrieval. To address the issue, this paper proposes a retrievable image encryption method based on GAN(RIE-GAN), which ensures the high security and good retrieval performance of the ciphertext. RIE-GAN uses the convolutional neural network to extract image feature and utilizes the trained weight vector from the Group Normalization (GN) layer to evaluate the importance of each channel for retrieval and divide the feature into two subsets of different importance, the important subset for retrieval and the non-important subset. To achieve similarity retrieval of the encrypted important subset, we utilize Variable Feature Thumbnail Preserving Encryption (VF-TPE) to ensure that the mean value within a block of the encrypted important subset remains unchanged for retrieval. To further enhance security, we use a three-dimensional Lorenz chaotic system to encrypt the non-important subset. The two encrypted subsets are then merged into a whole, serving as the target domain for Cycle-GAN. By training the Cycle-GAN, we perform the feature transformation from the source domain to the target domain, thus ensuring both the security and retrieval functionality of the encrypted data. Experimental results on the Corel-1000 dataset demonstrate that our method can reach 0.886 in the accuracy of ciphertext retrieval, significantly outperforming other method and the security of the ciphertext is close to that of other encryption methods.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"163 ","pages":"Article 105202"},"PeriodicalIF":2.9,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143769179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Contrastive activation maps with superpixel rectification for weakly supervised semantic segmentation: When superpixels meet CAMs

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-27 DOI: 10.1016/j.dsp.2025.105196

Huilin Shi, Yukun Liu, Shaofan Wang, Yanfeng Sun, Baocai Yin

{"title":"Contrastive activation maps with superpixel rectification for weakly supervised semantic segmentation: When superpixels meet CAMs","authors":"Huilin Shi, Yukun Liu, Shaofan Wang, Yanfeng Sun, Baocai Yin","doi":"10.1016/j.dsp.2025.105196","DOIUrl":"10.1016/j.dsp.2025.105196","url":null,"abstract":"<div><div>The weakly supervised semantic segmentation (WSSS) task amounts to segmenting all pixels by using weaker annotations instead of pixel-level ones. It suffers from two ubiquitous issues: over-activation and under-activation, incurred from unsatisfactory class activation maps (CAMs). Existing methods cannot balance this dilemma due to two reasons. (a) Most of methods learn less discriminative parts iteratively, leading to discontinuous parts of activation regions. (b) Most of methods refine CAMs by learning inter-pixel affinity from a global perspective which takes high complexity. We propose <u>C</u>ontrastive <u>A</u>ctivation <u>M</u>aps with <u>S</u>uperpixel <u>R</u>ectification (SRCAM) for WSSS by incorporating two effective tools: superpixel segmentation and contrastive learning with CAMs together. Typically, superpixels rectify those incomplete vanilla CAMs by examining each activation region based on its proportion with respect to each superpixel. After learning CAMs, SRCAM disentangles foreground objects and background from local patches by learning a cross-patch foreground-background contrast, and then preserves the invariance of activation maps from patch level to image level by learning a local-to-global activation map contrast. Experiments show that SRCAM achieves the state-of-the-art performance (72.8% mIoU) on the PASCAL VOC 2012 among 54 WSSS methods, and achieves the top-1 performance (44.8% mIoU) on the MS COCO 2014 among 27 WSSS methods. Code is available at <span><span>https://github.com/wangsfan/SRCAM</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"162 ","pages":"Article 105196"},"PeriodicalIF":2.9,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Few-shot specific emitter identification via asymmetric dual-path masked autoencoder

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-27 DOI: 10.1016/j.dsp.2025.105201

Sen Yang , Shui Yu , Qi Li , Keyang Xia , Hongna Zhu

{"title":"Few-shot specific emitter identification via asymmetric dual-path masked autoencoder","authors":"Sen Yang , Shui Yu , Qi Li , Keyang Xia , Hongna Zhu","doi":"10.1016/j.dsp.2025.105201","DOIUrl":"10.1016/j.dsp.2025.105201","url":null,"abstract":"<div><div>Specific emitter identification (SEI) occupies a pivotal position in device management and communication security. Due to hardware variations, each radiation source produces a distinctive radio frequency fingerprint, acting as a fundamental means for signal feature extraction in deep learning based on SEI. Traditional methods necessitate substantial labeled data, which are expensive and challenging. However, we propose a few-shot SEI (FS-SEI) method, based on an asymmetric dual-path masked autoencoder framework (ADPMAE). Especially, the encoder architecture incorporates two asymmetric branches, the main branch utilizing a time attention mechanism that integrating masking block to accurately capture local features of masked data. The other serves as the auxiliary branch and is applied to extracting global features of unmasked data. During the fine-tuning phase, a center loss function is introduced to optimize the encoder, which further enhances the identification performance. To evaluate the effectiveness of the proposed approach, we conduct several experiments on large-scale automatic dependent surveillance-broadcast (ADS-B) dataset and Wi-Fi dataset. The experiment results demonstrate that the ADPMAE outperforms other FS-SEI methods. Note that it can achieve 97.29 % identification accuracy, when 10 samples per ADS-B device. In addition, ADPMAE obtains an accuracy of 95.75 % under an identification task of Wi-Fi data.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"163 ","pages":"Article 105201"},"PeriodicalIF":2.9,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143746769","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Adaptive Poisson multi-Bernoulli filter for multiple extended targets with Gamma and Beta estimator

IF 2.9 3区工程技术

Digital Signal Processing Pub Date : 2025-03-27 DOI: 10.1016/j.dsp.2025.105204

Cheng Chen, Jinlong Yang, Jianjun Liu

引用次数: 0