Focal cortical dysplasia (FCD) is a developmental disorder frequently linked to drug-resistant focal epilepsy, where surgical intervention is often the most promising treatment. Unfortunately, conventional neuroradiological methods often struggle to detect subtle FCD cases, which may result in missed surgical opportunities for patients. To address this challenge, we have developed an innovative approach to FCD lesions detection using conditional diffusion models guided by complementary location-based modal factors. Our method involves employing a classifier to identify epileptic sites while directing the diffusion model to generate pseudo-healthy images. Using the unique features of FCD present in both T1 and FLAIR images, T1 images conditions the diffusion model to produce the corresponding FLAIR image, effectively removing abnormal tissue. Recognizing the difficulty in detecting epileptic lesions due to their subtle presentation, we have incorporated histogram matching techniques to address the color distortion issues commonly associated with diffusion models. This adjustment ensures that chromatic aberration does not hinder the identification of lesions. The effectiveness of our method has been validated using the UHB FCD MRI dataset, achieving an image-level recall metric of 0.952 and a pixel-level dice metric of 0.245. These results surpass those obtained from four other comparative methods, underscoring the superior performance of our approach. Our code is available at https://github.com/CodePYJ/FCD-Detection.

Diffusion models have revealed powerful potential in all-in-one image restoration (AiOIR), which is talented in generating abundant texture details. The existing AiOIR methods either retrain a diffusion model or fine-tune the pretrained diffusion model with extra conditional guidance. However, they often suffer from high inference costs and limited adaptability to diverse degradation types. In this paper, we propose an efficient AiOIR method, Diffusion Once and Done (DOD), which aims to achieve superior restoration performance with only one-step sampling of Stable Diffusion (SD) models. Specifically, multi-degradation feature modulation is first introduced to capture different degradation prompts with a pretrained diffusion model. Then, parameter-efficient conditional low-rank adaptation integrates the prompts to enable the fine-tuning of the SD model for adapting to different degradation types. Besides, a high-fidelity detail enhancement module is integrated into the decoder of SD to improve structural and textural details. Experiments demonstrate that our method outperforms existing diffusion-based restoration approaches in both visual quality and inference efficiency.

Unsupervised image super-resolution offers distinct advantages for real-world applications by eliminating the need for paired high- and low-resolution images. This paper proposes a novel architecture specifically designed for unsupervised learning, consisting of a cycle branch and a diffusion branch. The cycle branch integrates an upsampling and a downsampling network to generate pseudo-paired images from unpaired high- and low-resolution inputs. In parallel, the diffusion branch incorporates two independent diffusion models that refine these pseudo pairs, jointly modeling the processes of image reconstruction and degradation. This collaborative design enhances the authenticity of the pseudo pairs and enriches the detail in the reconstructed images. A key challenge in unsupervised learning is the lack of explicit label supervision, which often leads to inaccurate color restoration. To address this, we introduce a color consistency loss that regulates the cycle branch and promotes color fidelity. Through joint end-to-end training, the two branches complement each other to achieve high-quality reconstruction. Experimental results demonstrate that the proposed method effectively handles real-world low-resolution images, providing a robust and practical solution for image super-resolution.

Biomedical data exhibit high-dimensional complexity in its internal structure and are susceptible to noise interference, making classification tasks in biomedical data highly challenging. Twin support vector machine (TSVM) is a machine learning algorithm that can effectively solve pattern recognition problems. To mitigate the negative impact of noise, researchers have combined fuzzy set theory with TSVM and use fuzzy membership to describe the influence of different samples on constructing the optimal hyperplane, thus, extending TSVM to fuzzy twin support vector machines (FTSVM). In this paper, the dissimilarity measure based on data distribution is innovatively introduced into the fuzzy membership assignment process, and a novel fuzzy membership assignment strategy is designed to effectively reduce the negative impact of noise in biomedical data. Rather than rely on geometric distance, this strategy takes data distribution as the primary factor in measuring dissimilarity between samples and then constructs a heuristic function to assign fuzzy membership to different samples. Combining this strategy with TSVM, this paper proposed a fuzzy twin support vector machine based on dissimilarity measure (DFTSVM), which could effectively solve the classification problem with noise and shows excellent generalization performance in biomedical data. Moreover, DFTSVM employs a coordinate descent strategy with shrinking by active set to reduce computational complexity, which significantly improves the training speed of the model. Experiments are conducted on 14 biomedical datasets to compare the performance of DFTSVM with 10 heterogeneous machine learning classification algorithms and four homology algorithms. The results demonstrate that DFTSVM outperforms other algorithms in terms of classification performance on biomedical data. It exhibits excellent generalization performance in noisy environments, and its advantages in terms of generalization performance and noise robustness become more prominent as the noise rate increases.

Transformer has shown remarkable performance improvements over convolutional neural network (CNN) in natural language processing and high-level vision tasks. However, its application in low-level vision tasks, such as single image super-resolution (SISR), is still under-explored. In this paper, we introduce an up-down iterative algorithm and design a residual down and up Transformer block (RDUTB) in the Transformer framework. Then we propose a network for SISR based on RDUTB, which can effectively reconstruct low resolution (LR) images. Furthermore, to address the increasing demand for SISR models that can run on low-end mobile devices, we simplify the proposed model structure and adopt a content-based early-stopping strategy in the proposed SISR model to reduce the parameters and accelerate the reconstruction process while maintaining high quality. Experimental results show that our proposed Transformer-based SISR network and its lightweight version achieve superior performance over both traditional CNN-based SISR methods and some of the latest Transformer-based SISR methods.

Single image super-resolution with diffusion probabilistic models (SRDiff) is a successful diffusion model for image super-resolution that produces high-quality images and is stable during training. However, due to the long sampling time, it is slower in the testing phase than other deep learning-based algorithms. Reducing the total number of diffusion steps can accelerate sampling, but it also causes the inverse diffusion process to deviate from the Gaussian distribution and exhibit a multimodal distribution, which violates the diffusion assumption and degrades the results. To overcome this limitation, we propose a fast SRDiff (FSRDiff) algorithm that integrates a generative adversarial network (GAN) with a diffusion model to speed up SRDiff. FSRDiff employs conditional GAN to approximate the multimodal distribution in the inverse diffusion process of the diffusion model, thus enhancing its sampling efficiency when reducing the total number of diffusion steps. The experimental results show that FSRDiff is nearly 20 times faster than SRDiff in reconstruction while maintaining comparable performance on the DIV2K test set.

Background: Domain adaptation (DA) techniques have emerged as a pivotal strategy in addressing the challenges of cross-subject classification. However, traditional DA methods are inherently limited by the assumption of a homogeneous space, requiring that the source and target domains share identical feature dimensions and label sets, which is often impractical in real-world applications. Therefore, effectively addressing the challenge of EEG classification under heterogeneous spaces has emerged as a crucial research topic. Methods: We present a comprehensive framework that addresses the challenges of heterogeneous spaces by implementing a cross-domain class alignment strategy. We innovatively construct a cross-encoder to effectively capture the intricate dependencies between data across domains. We also introduce a tailored class discriminator accompanied by a corresponding loss function. By optimizing the loss function, we facilitate the aggregation of features with corresponding classes between the source and target domains, while ensuring that features from non-corresponding classes are dispersed. Results: Extensive experiments were conducted on two publicly available EEG datasets. Compared to advanced methods that combine label alignment with transfer learning, our method demonstrated superior performance across five heterogeneous space scenarios. Notably, in four heterogeneous label space scenarios, our method outperformed the advanced methods by an average of 7.8%. Moreover, in complex scenarios involving both heterogeneous label spaces and heterogeneous feature spaces, our method outperformed the state-of-the-art methods by an average of 4.1%. Conclusions: This paper presents an efficient model for cross-subject EEG classification under heterogeneous spaces, which significantly addresses the challenges of EEG classification within heterogeneous spaces, thereby opening up new perspectives and avenues for research in related fields.

Deep learning-based methods excel at single image super resolution (SR), but struggle with multi-frame image SR due to their inability to effectively exploit the complementary information within low resolution (LR) images. Graph cuts remain effective in this context. A major challenge for graph cuts lies in the need for the energy function (EF) to adhere to specific regularization constraints. To meet these constraints, existing methods often rely on approximations that can degrade reconstruction quality. To minimize these negative effects, we design a filter to convert the EF into a standard form suitable for graph cuts and introduce a filter-based SR model with maximum accuracy. To fully utilize the complementary information within LR images, we also propose a LR pixel selection mechanism that selects and weights LR pixels in our model. Experimental results demonstrate the robustness of our model against noise and point spread function misestimation. Moreover, our model outperforms existing algorithms in reconstructing fine-grained details.

Blind super-resolution (SR) aims to restore real low-resolution (LR) images. However, most current methods focus on global uniform blur but neglect motion blur, and the few motion deblurring SR methods tend to produce too smooth images. In this letter, we introduce a novel diffusion-based SR method, which can effectively handle the motion blur effect in LR images and retain fine-grained texture information. Our method uses a deblurred feature extraction module and a texture feature extraction module to obtain deblurred features and texture features of the LR image respectively. These two features are then fed into the diffusion model, which samples the image from a learned distribution and outputs a clear and realistic HR image. Moreover, to speed up the sampling process of the diffusion model, we combine it with a conditional generative adversarial network (GAN) to implement stride sampling. Extensive experiments show that our method outperforms state-of-the-art methods in terms of perceptual metrics, and can generate more natural and realistic images.

Image super-resolution (SR) usually synthesizes degraded low-resolution images with a predefined degradation model for training. Existing SR methods inevitably perform poorly when the true degradation does not follow the predefined degradation, especially in the case of the real world. To tackle this robustness issue, we propose a cascaded degradation-aware blind super-resolution network (CDASRN), which not only eliminates the influence of noise on blur kernel estimation but also can estimate the spatially varying blur kernel. With the addition of contrastive learning, our CDASRN can further distinguish the differences between local blur kernels, greatly improving its practicality. Experiments in various settings show that CDASRN outperforms state-of-the-art methods on both heavily degraded synthetic datasets and real-world datasets.

分析表明,3D打印技术可以快速生成三维实体模型,对制造业有重要的价值,且应用领域十分广泛.阐述3D打印数据的传输面临复制、修改、传播等问题,使用AES算法对3D打印的数据进行加密.

Deep learning has made great contributions to the study of single image super resolution. The recently proposed feed-forward architectures of super-resolution focus on nonlinear mapping from low-resolution inputs to high-resolution outputs. However, the feed-forward structure does not well represent the interdependencies between low- and high-resolution images. This leads to bad effect of SISR for large scaling factor. To solve this problem, this paper proposes an enhanced back-projection network that provides an up and down sampling process with error feedback to capture various spatial correlations, and introduces the residual block in sampling process to alleviate the difficulty of training deep networks and achieve better results. The results about 8x SR show that the proposed network is effective with compare to other popular methods in the large scaling factor.

移动端的表情识别有巨大需求,但是受算力限制,主流深度神经网络无法直接移植.为此,设计了一个浅层网络,在节约计算量的同时保证了识别率.网络中使用三组堆叠而成的卷积层,有助于增大感受野,便于更好地提取特征,这是提升识别率的关键;使用全局平均池化层,避免引入额外的全连接层,大幅降低参数量,在训练样本不足的情况下,降低模型过拟合风险.在FER-2013数据集进行训练,准确率超过现有大多数算法;在CK+数据集上进行微调,测试集上的准确率可达到0.96.将所得模型转换为Core ML模型,结合Xcode平台在iOS端搭建了实时表情识别App,在iPhone 8 Plus上能够稳定、流畅运行,识别效果达到预期.

目的:解决智能舌诊中肝郁线自动识别问题.方法:设计了一个掩模,在使用深度卷积神经网络模型自动提取特征时,专注于肝郁线所在部位.在此特征基础上,搭建了一个前馈神经网络,用于识别肝郁线.通过随机调整舌体图像的方式增强数据,解决样本数量不平衡的问题.结果:掩模和数据增强可以大幅提升识别精确率,本研究所提模型的各项指标均优于经典分类算法.结论:在舌体图像清晰可辨的情况下,使用机器自动识别肝郁线是切实可行的.

MATLAB作为一款科学计算软件,除了强大的编程、计算、建模等功能外,还提供了用于制作GUI(Graphical User Interfaces,图形用户界面)的一系列接口,通过使用这些接口,可以很方便地创建出具有特定功能的GUI程序.本文使用GUIDE方法,通过设置控件的不同属性、编写对应的回调函数,并结合MATLAB的事件监听技术,设计并实现了一个具有视频录制、播放进度控制及三通道播放模式等功能的视频播放器.

In this study, we consider the use of the convolution method for constructing approximations comprising fuzzy number sequences with useful properties for a general fuzzy number. We show that this convolution method can generate differentiable approximations in finite steps for fuzzy numbers with finite non-differentiable points. In previous studies, this convolution method was only used for constructing differentiable approximations of continuous fuzzy numbers, the possible non-differentiable points of which were the two endpoints of the 1-cut. The construction of smoothers is a key step in the process for producing approximations. We also show that if appropriate smoothers are selected, then we can use the convolution method to provide approximations that are differentiable, Lipschitz, and that also preserve the core.

This paper proposed an improved Local Binary Features (LBF) [1] algorithm for bayonet surveillance system. Since LBF is based on shape-regression strategy, which is prone to over-fitting after multi-stage regression, the training model cannot be directly applied to other scenarios. To this end, we employed new data into existing models at the final stage of regression. As a consequent, newly imported data can be embedded to the model. The experimental results conducted on bayonet surveillance videos showed that the proposed method outperforms LBF, ERT and SDM, concerning with mean errors of the regression results.

In this letter, we propose a new graph cuts multi-frame super resolution method. The method is carried out in 3 steps. First, we project each high-resolution pixel p onto the low-resolution images and select low-resolution pixels which fall within the zone of influence of p. Second, we weigh the contribution of the low-resolution pixels via a soft switching function and add them to construct a virtual low resolution pixel. The high resolution image is then recovered after minimizing a Maximum a posteriori Markov Random Field (MAP-MRF) energy function. This is done by approximating our energy function to make it graph representable and minimize it with a graph cuts α-expansion algorithm. Experimental results show that our approach outperforms state-of-the-art methods.

This paper studies the problem of multi-frame image super-resolution reconstruction. The process of image degradation is modeled by using the first-order Taylor expansion based on sub-pixel. Then the energy minimization function is established and the graph-cut algorithm is chosen to solve the energy minimization. In order to confirm this algorithm, we obtain the low resolution images by two ways: simulating image degradation and taking photos. By comparing the 4 × 4 times reconstruction results, it is shown that this algorithm is valid not only for simulation of low resolution images but also for real images. Besides, experimental results show that this algorithm possesses good anti-interference ability of noise.

Image denoising is one of the key issues in the image processing and the foundation of further research.Combined with the sparse representation theory,which emerged in recent year,we can handle the image denoising problems better.Based on orthogonal matching pursuit(OMP) algorithm,this paper used K-singular value decomposition(K-SVD) algorithm for image de-noising.In order to get better de-noising performance,this paper improves dictionary updating algorithm.The core provides a more optimal choice for training of the dictionary atoms,replaces the useless and redundant dictionary of atoms with natural image patch dictionary atoms.By this way,we improve the training of the dictionary effectively,and adapt to natural image.Experimental results show that,compare with the wavelet de-noising algorithm,this algorithm has a good de-noising ability,while keeping the detail and the edge character of the image better,make the de-noising image clear.