Abstract Zero-Reference Deep Curve Estimation (Zero-DCE) is currently one of the most popular low-light image enhancement methods. Through extensive experimentation, we observe that: (i) excellent performance Zero-DCE depends on training data with multiple exposure levels, (ii) it cannot effectively handle uneven light, extremely low or overexposed images in natural environments. Therefore, propose an improved zero-reference dual-illumination deep curve estimation method for named...

Fusing classifiers’ decisions can improve the performance of a pattern recognition system. Many applications areas have adopted methods multiple classifier fusion to increase classification accuracy in process. From fully considering differences and training sample information, algorithm using weighted decision templates is proposed this paper. The uses statistical vector measure classifier’s makes weighed transform on each according reliability its output. To make decision, information...

Aiming at solving the difficulty of modeling on spatial coherence, complete feature extraction, and sparse representation in hyperspectral image classification, a joint classification method is investigated by flexible patches sampling superpixels. First, principal component analysis total variation diffusion are employed to form pseudo color for simplifying superpixels computing with (simple linear iterative clustering) SLIC model. Then, we design recovery model overcomplete estimate...

This paper applies pattern recognition approach based on classifier selection to network intrusion detection and proposes a clustering-based method. In the method, multiple clusters are selected for test sample. Then, average performance of each is calculated with best chosen classify calculation performance, weighted adopted. Weight values according distances between sample cluster. Experiments were done dataset KDD'99 compare method Clustering Selection (CS) The experimental results show...

Crowd counting is a task that aims to estimate the number of people in an image. Recent crowd methods make significant progress by employing convolutional neural networks regress density maps. One most challenging problems this drastic scale variation region interest images. In paper, Feature Fusion Attention Network (FFANet) proposed for counting. Firstly, VGG16 network adapted as backbone FFANet extract features Then, extracted are fused subsequent two stages. Specifically, information...

In this paper, a novel spectral-spatial joint sparse representation algorithm for hyperspectral image classification is proposed based on multi-layer superpixels in various scales.Superpixels of scales can provide complete yet redundant correlated information the class attribute test pixels.Therefore, we design model pixel by sampling similar pixels from its corresponding combinations.Firstly, are extracted false color HSI data principal components analysis model.Secondly, group...

<title>Abstract</title> Weakly supervised semantic segmentation (WSSS) using only image-level labels has gradually become an emerging research hotspot in the field of computer vision recent years due to its low annotation cost. Existing methods rely on Class Activation Maps (CAMs) from specific classification models locate target regions. However, classifiers tend focus most discriminative regions input image and assign higher weights these areas, leading problem incomplete CAM To address...

SoftMan, the virtual robot in network environment, is a kind of software Artificial Life living computer networks. It has humanoid structure, and simulating human, its perception system should be able to recognize perceptive objects. Enlightened by human's "disassemble-integration" method analyzing large systems, cooperative classification model SoftMan's proposed. In model, different senses are simulated multiple classifier systems (MCS). Consequently, SoftMan can perform multi-sense on A...

At present, fixed rules for classifier fusion are the most used and widely investigated ones, which, with no second- level training, compete more sophisticated rules. But, one problem is, that although they have good overall performance, it is not clear which a particular data set. In this paper, an experimental comparison of well-known six (product, mean, maximum, minimum, median majority voting) was done on some sets KDD'99, UCI ELENA. The results allow draw preliminary conclusions about...

Currently, Multiple Classifier System (MCS) attracts more and attentions has become one of the research hotspots in pattern recognition field. selection is a commonly used strategy for MCS to achieve final decision. A classifier method based on clustering w eighted mean proposed this paper. In method, multiple clusters are selected according distances between cluster centers input sample. Then, average performance each calculated. The best nearest with picked out. According reliability their...

Ideal color image segmentation needs both low-level cues and high-level semantic features. This paper proposes a two-hierarchy model based on merging homogeneous superpixels. First, region growing strategy is designed for producing homogenous compact superpixels in different partitions. Total variation smoothing features are adopted the procedure locating real boundaries. Before merging, we define combined color-texture histogram feature description and, meanwhile, novel objectness proposed...

SoftMan, the virtual robot in network environment, is a kind of software Artificial Life living computer networks. It has humanoid structure, and simulating human, its perception system should be able to recognize perceptive objects. This paper proposes cooperative classification model for SoftMan's system, which, different senses are simulated by multiple classifier systems (MCS). An essential issue research MCS method combination. Accordingly, this combination (called CMS), which improves...

Abstract: Haze rendering aims to generate realistic nighttime haze images from clear images. The results can be applied various practical applications, such as image dehazing algorithms, game scene rendering, shooting filters, etc. We investigate two smaller but challenging problems in namely 1) how accurately estimate the transmission map and air light images, respectively, absence of paired datasets? 2) How render characteristics a image: glow effect? For this purpose, we propose an...

Abstract: Most low-light enhancement methods based on curve estimation do not have the ability of multi-illumination processing. They cannot use one model to process images with different exposure levels. This paper proposes by dynamic estimation. The is a new light that allows appropriate enhancement, suppression, and hold operations depending level input image. Therefore, can in lighting environments. In addition, further improve model's illumination images, discriminator designed. It...

Abstract Real‐time semantic segmentation is a crucial technology in automatic driving scenarios, which needs to meet both high precision and real‐time. The authors observe that learning complex correlations between object categories vital the real‐time task. Moreover, image spatial detail information plays an important role small preserving edges textures. A Semantics Recalibration Detail Enhancement Network for based on BiSeNet V2 proposed. On one hand, lightweight module designed...

Clustering and Selection (CS) is a common method of multiple classifier selection. But the judging an input sample belong to certain area just by shortest distance has some unilateralism. Therefore, dual selection based on clustering proposed. In method, clusters are selected for test with best weighted average performance chosen. The chosen compared in nearest cluster better one used classify sample. main parameter self-adaptively according prior information training samples. Experiments...

SoftMan, the virtual robot in network environment, is a kind of software artificial life living computer networks. It has humanoid structure, and simulating human, its perception system should be able to recognize perceptive objects. Enlightened by human's "disassemble-integration" method analyzing large systems, cooperative classification model SoftMan's proposed. In model, different senses are simulated multiple classifier systems (MCS). An essential issue research MCS combination....