Global Refinement Research Articles

Spectrum Resource Utilization and Security Enhancement in Multi-domain Elastic Optic Networks (MD-EON) Using Memetic Algorithm

Routing and Spectrum utilization is a challenging approach when optical signal is used in data transfer. The underlying bandwidth requirements in supporting the optical signal transfer and appropriate fragmentation are to be considered in routing and spectrum allocation. The problem lies with assigning a large bandwidth for a shorter duration of time or smaller bandwidth is provisioned for a larger duration. Thus the data transfer using memetic algorithm-elastic optical network (MA-EON) incorporates suggestion for crossing from one elastic optical network (EON) to different EON with its current spectrum utilization termed as global maximum. This flow within the EON is calculated as local refinement and is improved subsequently to obtain global refinement. MA-EON based finds better performance in terms of computational time in finding diverse routes and reduced blocking probability when compared to existing system used in this work.

Building outline delineation: From aerial images to polygons with an improved end-to-end learning framework

Deep learning methods based upon convolutional neural networks (CNNs) have demonstrated impressive performance in the task of building outline delineation from very high resolution (VHR) remote sensing (RS) imagery. In this paper, we introduce an improved method that is able to predict regularized building outline in a vector format within an end-to-end deep learning framework. The main idea of our framework is to learn to predict the location of key vertices of the buildings and connect them in sequence. The proposed method is based on PolyMapper. We upgrade the feature extraction by introducing global context and boundary refinement blocks and add channel and spatial attention modules to improve the effectiveness of the detection module. In addition, we introduce stacked conv-GRU to further preserve the geometric relationship between vertices and accelerate inference. We tested our method on two large-scale VHR-RS building extraction dataset. The results on both COCO and PoLiS metrics demonstrate better performance compared with Mask R-CNN and PolyMapper. Specifically, we achieve 4.2 mask mean average precision (mAP) and 3.7 mean average recall (mAR) absolute improvements compared to PolyMapper. Also, the qualitative comparison shows that our method significantly improves the instance segmentation of buildings of various shapes.

LPAQR-Net: Efficient Vertebra Segmentation From Biplanar Whole-Spine Radiographs.

Vertebra segmentation from biplanar whole-spine radiographs is highly demanded in the quantitative assessment of scoliosis and the resultant sagittal deformities. However, automatic vertebra segmentation from the radiographs is extremely challenging due to the low contrast, blended boundaries, and superimposition of many layers, especially in the sagittal plane. To alleviate these problems, we propose a lightweight pyramid attention quick refinement network (LPAQR-Net) for efficient and accurate vertebra segmentation. The LPAQR-Net consists of three components: (1) a lightweight backbone network (LB-Net) to prune network parameters and memory footprints to strike an optimal balance between speed and accuracy, (2) a series of global attention refinement (GAR) modules to selectively reuse low-level features to facilitate the feature refinement, and (3) an attention-based atrous spatial pyramid pooling (A-ASPP) module to extract weighted pyramid contexts to improve the segmentation of blurred vertebrae. Moreover, the multi-class training strategy is employed to alleviate the over-segmentation of adjacent vertebrae. Evaluation results on both frontal and lateral radiographs of 332 AIS patients show our method achieves accurate vertebra segmentation with significant reductions in inference time and computational demands compared to the state-of-the-art. Meanwhile, results on the public AASCE2019 dataset also demonstrate the good generalization ability of our model. It is the first attempt to explore the lightweight network for vertebra segmentation from biplanar whole-spine radiographs. It simulates radiologists gathering nearby contexts for accurate and robust vertebra boundary inference. The method can provide efficient and accurate vertebra segmentation for clinicians to perform a fast and reproducible spinal deformity evaluation.

Deep Localization of Static Scans in Mobile Mapping Point Clouds

Mobile laser scanning (MLS) systems are often used to efficiently acquire reference data covering a large-scale scene. The terrestrial laser scanner (TLS) can easily collect high point density data of local scene. Localization of static TLS scans in mobile mapping point clouds can afford detailed geographic information for many specific tasks especially in autonomous driving and robotics. However, large-scale MLS reference data often have a huge amount of data and many similar scene data; significant differences may exist between MLS and TLS data. To overcome these challenges, this paper presents a novel deep neural network-based localization method in urban environment, divided by place recognition and pose refinement. Firstly, simple, reliable primitives, cylinder-like features were extracted to describe the global features of a local urban scene. Then, a probabilistic framework is applied to estimate a similarity between TLS and MLS data, under a stable decision-making strategy. Based on the results of a place recognition, we design a patch-based convolution neural network (CNN) (point-based CNN is used as kernel) for pose refinement. The input data unit is the batch consisting of several patches. One patch goes through three main blocks: feature extraction block (FEB), the patch correspondence search block and the pose estimation block. Finally, a global refinement was proposed to tune the predicted transformation parameters to realize localization. The research aim is to find the most similar scene of MLS reference data compared with the local TLS scan, and accurately estimate the transformation matrix between them. To evaluate the performance, comprehensive experiments were carried out. The experiments demonstrate that the proposed method has good performance in terms of efficiency, i.e., the runtime of processing a million points is 5 s, robustness, i.e., the success rate of place recognition is 100% in the experiments, accuracy, i.e., the mean rotation and translation error is (0.24 deg, 0.88 m) and (0.03 deg, 0.06 m) on TU Delft campus and Shanghai urban datasets, respectively, and outperformed some commonly used methods (e.g., iterative closest point (ICP), coherent point drift (CPD), random sample consensus (RANSAC)-based method).

Adaptive Extended Isogeometric Analysis for Steady-State Heat Transfer in Heterogeneous Media

Steady-state heat transfer problems in heterogeneous solid are simulated by developing an adaptive extended isogeometric analysis (XIGA) method based on locally refined non-uniforms rational B-splines (LR NURBS). In the XIGA, the LR NURBS, which have a simple local refinement algorithm and good description ability for complex geometries, are employed to represent the geometry and discretize the field variables; and some special enrichment functions are introduced into the approximation of temperature field, thus the computational mesh is independent of the material interfaces, which are described with the level set method. Similar to the approximation of temperature field, a temperature gradient recovery technique for heterogeneous media is proposed, and based on the Zienkiewicz–Zhu recovery technique a posteriori error estimator is defined to automatically identify the locally refined regions. The convergence and performance properties of the developed method are verified by using three numerical examples. The numerical results show that (1) The convergence speed of the adaptive local refinement is faster than that of the uniform global refinement; (2) The convergence rate of the high-order basis functions is faster than that of the low-order basis functions; and (3) The existing inclusions change the local distributions of the temperature, and the extreme values of the temperature gradients take place around the inclusion interfaces.

Hybrid grey wolf sunflower optimisation algorithm for energy‐efficient cluster head selection in wireless sensor networks for lifetime enhancement

Wireless sensor networks (WSNs) are expected to find extensive applicability and accelerating deployment in the future. However, the main challenge faced in WSN is its perishing lifetime. The process of clustering a network is a popular mechanism employed for the purpose of extending the lifespan of WSNs and thereby making efficient data transmission. The main aim of a clustering algorithm is to elect an optimal cluster head (CH). The recent research trend suggests meta-heuristic algorithms for the selection of optimal CHs. Meta-heuristic algorithms possess the advantages of being simple, flexible, derivation-free, and avoids local optima. This research proposes a novel hybrid grey wolf optimiser-based sunflower optimisation (HGWSFO) algorithm for optimal CH selection (CHS) under certain factor constraints such as energy spent and separation distance, such that the network lifetime is enhanced. Sunflower optimisation (SFO) is employed for a broader search (exploration) where the variation of the step-size parameter brings the plant closer to the sun in search of global refinement, thus increasing the exploration efficiency. Grey wolf optimisation (GWO) is employed for a narrow search (exploitation), where the parameter coefficient vectors are deliberately required to emphasise exploitation. This balances the exploration-exploitation trade-off, prolongs the network lifetime, increases the energy efficiency, and enhances the performance of the network with respect to overall throughput, residual energy of nodes, dead nodes, alive nodes, network survivability index, and convergence rate. The superior characteristic of the suggested HGWSFO is validated by comparing its performance with various other existing CHS algorithms. The overall performance of the proposed HGWSFO is 28.58%, 31.53%, 48.8%, 49.67%, 54.95%, 70.76%, and 87.10%, better than that of GWO, SFO, particle swarm optimisation (PSO), improved PSO, low-energy adaptive clustering hierarchy (LEACH), LEACH-centralised, and direct transmission, respectively.

Comparing Q-Tree with Nested Grids for Simulating Managed River Recharge of Groundwater

The use of rivers to recharge groundwater is a key water resource management method. High-precision simulations of the groundwater level near rivers can be used to accurately assess the recharge effect. In this study, we used two unstructured grid refinement methods, namely, the quadtree (Q-tree) and nested grid refinement techniques, to simulate groundwater movement under river recharge. We comparatively analyzed the two refinement methods by considering the simulated groundwater level changes before and after the recharge at different distances from the river and by analyzing the groundwater flow and model computation efficiency. Compared to the unrefined model, the two unstructured grid refinement models significantly improve the simulation precision and more accurately describe groundwater level changes from river recharge. The unstructured grid refinement models have higher calculation efficiencies than the base model (the global refinement model) without compromising the simulation precision too much. The Q-tree model has a higher simulation precision and a lower computation time than the nested grid model. In summary, the Q-tree grid refinement method increases the computation efficiency while guaranteeing simulation precision at a certain extent. We therefore recommended the use of this grid refinement method in simulating river recharge to the aquifers.

Three-dimensional direct current resistivity forward modeling based on the hp-adaptive finite element method

In recent years, the adaptive finite element method has been used for direct current resistivity forward modeling to improve the accuracy of numerical solutions. The accuracy of adaptive finite element solutions is mainly affected by two factors: the cell size (h) and the order of the shape function (p). To further improve the accuracy of the adaptive finite element solutions and keep low computational costs, we present a hp-adaptive finite element algorithm, combining h-adaptive and p-adaptive, for three-dimensional direct current forward modeling. The adaptive mesh refinement is guided by a posteriori error estimator and a smoothness estimator of the solution. In the adaptive process, for a cell with a large error, if the smoothness of the solution is low, the cell will be refined; otherwise, the order of the shape function on the cell will be increased by one order. To obtain high-precision finite element solutions for complex models with topographies the unstructured grids are adopted. The octree-based mesh refinement method is used to refine the cells, and the shape functions with arbitrary orders in three-dimensional space are generated by using tensor products of one-dimensional polynomials. A two-layer model is used to verify the correctness of the algorithm. The comparison between the convergence rate of global, h-adaptive, and hp-adaptive refinement algorithms indicates that our algorithm can provide the most accurate solution with the lowest computation costs. It shows exponentially convergence rate. Finally, a topography model with an abnormal body and a complex abnormal model are used to verify the robustness of our algorithm.

RefineU-Net: Improved U-Net with progressive global feedbacks and residual attention guided local refinement for medical image segmentation

Abstract Motivated by the recent advances in medical image segmentation using a fully convolutional network (FCN) called U-Net and its modified variants, we propose a novel improved FCN architecture called RefineU-Net. The proposed RefineU-Net consists of three modules: encoding module (EM), global refinement module (GRM) and local refinement module (LRM). EM is backboned by pretrained VGG-16 using ImageNet. GRM is proposed to generate intermediate layers in the skip connections in U-Net. It progressively upsamples the top side output of EM and fuses the resulted upsampled features with the side outputs of EM at each resolution level. Such fused features combine the global context information in shallow layers and the semantic information in deep layers for global refinement. Subsequently, to facilitate local refinement, LRM is proposed using residual attention gate (RAG) to generate discriminative attentive features to be concatenated with the decoded features in the expansive path of U-Net. Three modules are trained jointly in an end-to-end manner thereby both global and local refinement are performed complementarily. Extensive experiments conducted on four public datasets of polyp and skin lesion segmentation show the superiority of the proposed RefineU-Net to multiple state-of-the-art related methods.

Adaptive isogeometric analysis for phase‐field modeling of anisotropic brittle fracture

SummaryThe surface energy a phase‐field approach to brittle fracture in anisotropic materials is also anisotropic and gives rise to second‐order gradients in the phase field entering the energy functional. This necessitates C1 continuity of the basis functions which are used to interpolate the phase field. The basis functions which are employed in isogeometric analysis (IGA), such as nonuniform rational B‐splines and T‐splines naturally possess a higher order continuity and are therefore ideally suited for phase‐field models which are equipped with an anisotropic surface energy. Moreover, the high accuracy of spline discretizations, also relative to their computational demand, significantly reduces the fineness of the required discretization. This holds a fortiori if adaptivity is included. Herein, we present two adaptive refinement schemes in IGA, namely, adaptive local refinement and adaptive hierarchical refinement, for phase‐field simulations of anisotropic brittle fracture. The refinement is carried out using a subdivision operator and exploits the Bézier extraction operator. Illustrative examples are included, which show that the method can simulate highly complex crack patterns such as zigzag crack propagation. An excellent agreement is obtained between the solutions from global refinement and adaptive refinement, with a reasonable reduction of the computational effort when using adaptivity.

A locally refined adaptive isogeometric analysis for steady-state heat conduction problems

This paper presents an isogeometric analysis with adaptivity using locally refined B-splines (LR B-splines) for steady-state heat conduction simulations in solids. Within this framework, the LR B-splines, which have an efficient and simple local refinement algorithm, are used to represent the geometry, and are also employed for spatial discretization, thus providing a seamless interaction between the CAD models and the numerical analysis. A Zienkiewicz-Zhu a posteriori error estimator in terms of the temperature gradient recovery is used to identify the regions for local mesh refinement. The accuracy and convergence properties of the proposed framework are demonstrated through several two-dimensional isotropic examples. Numerical results indicate good performance of the present method as the adaptive refinement technique yields more accurate solution compared to the uniform global refinement with an improved convergence rate.

Optical flow refinement using iterative propagation under colour, proximity and flow reliability constraints

This study proposes a strategy to refine optical flow based on the estimated reliability maps. These maps are firstly estimated a posteriori after the motion estimation by the well-known Kanade–Lucas–Tomasi (KLT). With two new defined criteria based, respectively, on the optical flow local variance and the temporal evolution of the KLT residuals, a global refinement of the motion map is then carried out through two stages under the control of the reliability measures and the colour local homogeneousness. According to the experiments performed on the Middlebury dataset, the authors' reliability measures prove to be a good indicator for the quality of the estimation. Indeed, the correction process increases the global reliability measures and reduces the global errors in a significant way. The experiments show that the quality is higher than classical estimation methods and ranked at 88/168 on Middlebury website

Patch Proposal Network for Fast Semantic Segmentation of High-Resolution Images

Despite recent progress on the segmentation of high-resolution images, there exist an unsolved problem, i.e., the trade-off among the segmentation accuracy, memory resources and inference speed. So far, GLNet is introduced for high or ultra-resolution image segmentation, which has reduced the computational memory of the segmentation network. However, it ignores the importances of different cropped patches, and treats tiled patches equally for fusion with the whole image, resulting in high computational cost. To solve this problem, we introduce a patch proposal network (PPN) in this paper, which adaptively distinguishes the critical patches from the trivial ones to fuse with the whole image for refining segmentation. PPN is a classification network which alleviates network training burden and improves segmentation accuracy. We further embed PPN in a global-local segmentation network, instructing global branch and refinement branch to work collaboratively. We implement our method on four image datasets:DeepGlobe, ISIC, CRAG and Cityscapes, the first two are ultra-resolution image datasets and the last two are high-resolution image datasets. The experimental results show that our method achieves almost the best segmentation performance compared with the state-of-the-art segmentation methods and the inference speed is 12.9 fps on DeepGlobe and 10 fps on ISIC. Moreover, we embed PPN with the general semantic segmentation network and the experimental results on Cityscapes which contains more object classes demonstrate the generalization ability on general semantic segmentation.

Global Registration of Terrestrial Laser Scanner Point Clouds Using Plane-to-Plane Correspondences

Registration of point clouds is a central problem in many mapping and monitoring applications, such as outdoor and indoor mapping, high-speed railway track inspection, heritage documentation, building information modeling, and others. However, ensuring the global consistency of the registration is still a challenging task when there are multiple point clouds because the different scans should be transformed into a common coordinate frame. The aim of this paper is the registration of multiple terrestrial laser scanner point clouds. We present a plane-based matching algorithm to find plane-to-plane correspondences using a new parametrization based on complex numbers. The multiplication of complex numbers is based on analysis of the quadrants to avoid the ambiguity in the calculation of the rotation angle formed between normal vectors of adjacent planes. As a matching step may contain several matrix operations, our strategy is applied to reduce the number of mathematical operations. We also design a novel method for global refinement of terrestrial laser scanner data based on plane-to-plane correspondences. The rotation parameters are globally refined using operations of quaternion multiplication, while the translation parameters are refined using the parameters of planes. The global refinement is done non-iteratively. The experimental results show that the proposed plane-based matching algorithm efficiently finds plane correspondences in partial overlapping scans providing approximate values for the global registration, and indicate that an accuracy better than 8 cm can be achieved by using our global fine plane-to-plane registration method.

SpikeSegNet-a deep learning approach utilizing encoder-decoder network with hourglass for spike segmentation and counting in wheat plant from visual imaging

BackgroundHigh throughput non-destructive phenotyping is emerging as a significant approach for phenotyping germplasm and breeding populations for the identification of superior donors, elite lines, and QTLs. Detection and counting of spikes, the grain bearing organs of wheat, is critical for phenomics of a large set of germplasm and breeding lines in controlled and field conditions. It is also required for precision agriculture where the application of nitrogen, water, and other inputs at this critical stage is necessary. Further, counting of spikes is an important measure to determine yield. Digital image analysis and machine learning techniques play an essential role in non-destructive plant phenotyping analysis.ResultsIn this study, an approach based on computer vision, particularly object detection, to recognize and count the number of spikes of the wheat plant from the digital images is proposed. For spike identification, a novel deep-learning network, SpikeSegNet, has been developed by combining two proposed feature networks: Local Patch extraction Network (LPNet) and Global Mask refinement Network (GMRNet). In LPNet, the contextual and spatial features are learned at the local patch level. The output of LPNet is a segmented mask image, which is further refined at the global level using GMRNet. Visual (RGB) images of 200 wheat plants were captured using LemnaTec imaging system installed at Nanaji Deshmukh Plant Phenomics Centre, ICAR-IARI, New Delhi. The precision, accuracy, and robustness (F1 score) of the proposed approach for spike segmentation are found to be 99.93%, 99.91%, and 99.91%, respectively. For counting the number of spikes, “analyse particles”—function of imageJ was applied on the output image of the proposed SpikeSegNet model. For spike counting, the average precision, accuracy, and robustness are 99%, 95%, and 97%, respectively. SpikeSegNet approach is tested for robustness with illuminated image dataset, and no significant difference is observed in the segmentation performance.ConclusionIn this study, a new approach called as SpikeSegNet has been proposed based on combined digital image analysis and deep learning techniques. A dedicated deep learning approach has been developed to identify and count spikes in the wheat plants. The performance of the approach demonstrates that SpikeSegNet is an effective and robust approach for spike detection and counting. As detection and counting of wheat spikes are closely related to the crop yield, and the proposed approach is also non-destructive, it is a significant step forward in the area of non-destructive and high-throughput phenotyping of wheat.

A large‐scale parallel hybrid grid generation technique for realistic complex geometry

SummaryHigh‐Performance Computing (HPC) systems and Computational Fluid Dynamics (CFD) have made significant progress in recent years; however, as the basis of the large‐scale parallel computing, the massive grid generation of billions of cells has become a bottleneck problem. In this study, a parallel grid generation technique is proposed to generate large‐scale mixed grids with arbitrary cell types and scales. The basic idea of our method is analogous to the global mesh refinement technique. An initial coarse grid with arbitrary cell types is regarded as a background mesh which is partitioned into subzones, and subzones are assigned onto different CPU cores. After the cells and faces in each subzone are split, the inserted new points of the solid wall are projected onto the original CAD entities to preserve the geometry accurately. Finally, the tangled cells caused by the projection in the boundary layer are untangled by a local Radial Basis Function mesh deformation technique. Furthermore, a parallel partition approach and an efficient wall distance computing technique for massive grids are developed also to shorten the preprocessing time. The tests show that the preprocessing efficiency has been increased by two or three orders compared with traditional methods. Billions of grids are generated for the AIAA JSM high‐lift model and the Chinese CHN‐T1 transport model to test the ability of the parallel grid generation technique. The maximum scale up to 19 billion mixed elements is generated using 16 384 CPU cores in parallel, and the mesh quality is acceptable for CFD simulations.

Moving mesh version of wave propagation algorithm based on augmented Riemann solver

Shallow water equations are a nonlinear system of hyperbolic conservation laws. This system admits discontinuous solutions. In addition, a discontinuous bathymetry can lead to the discontinuous solutions. The accuracy of numerical methods depends on the accuracy of the numerical solutions around these discontinuities.In order to increase the accuracy, the adaptive mesh methods are more cost-effective than the methods based on the global refinement, In terms of computational cost. Moving mesh method is an adaptive mesh method. In this method, an increase of the number of mesh points is not needed for increasing the accuracy and efficiency of numerical solutions. Highly accurate solutions only can be obtained by concentrating points in areas where more accuracy is needed and decreasing the number of points in smooth areas.In this paper, LeVeque wave propagation algorithm is improved by the moving mesh method, for shallow water equations with variable bathymetry. Moreover, augmented Riemann solver is used for solving Riemann problems in each time step. Finally, moving mesh version of wave propagation algorithm based on augmented Riemann solver is proposed. The numerical solutions based on the moving mesh method and fixed mesh method have a significant difference and application of a moving mesh method yields highly accurate solutions around shocks and discontinuities.

Truncated hierarchical B-spline–based topology optimization

This work presents a truncated hierarchical B-spline–based topology optimization (THB-TO) to address topology optimization (TO) for both minimum compliance and compliant mechanism. The sensitivity and density filters with a lower bound are adaptively consistent with the hierarchical levels of active elements to remove the checkboard pattern and reduce the gray transition area. By means of the maximum variation of design variables on two consecutive iterative steps and the density differences of adjacent active elements, a mark strategy is established, which triggers the hierarchical local refinement and identifies the elements to be refined during the course of THB-TO. Besides, a locally refined design space is constructed in terms of the parent–child relationship of the cells on consecutive hierarchical levels. Numerical examples are used to verify the effectiveness of the proposed THB-TO, where the resolution around the boundary of the optimized designs can be effectively improved by THB-TO. Compared with global refinement, the number of degree of freedoms (DOFs) and design variables are largely decreased for 2D and 3D cases by THB-TO, which demonstrates that the proposed THB-TO is a promising approach to solving 2D and 3D TO problems.

Performance Comparison of Geoid Refinement between XGM2016 and EGM2008 Based on the KTH and RCR Methods: Jilin Province, China

The selection of an appropriate global gravity field model and refinement method can effectively improve the accuracy of the refined regional geoid model in a certain research area. We analyzed the accuracy of Experimental Geopotential Model (XGM2016) based on the GPS-leveling data and the modification parameters of the global mean square errors in the KTH geoid refinement in Jilin Province, China. The regional geoid was refined based on Earth Gravitational Model (EGM2008) and XGM2016 using both the Helmert condensation method with an RCR procedure and the KTH method. A comparison of the original two gravity field models to the GPS-leveling benchmarks showed that the accuracies of XGM2016 and EGM2008 in the plain area of Jilin Province are similar with a standard deviation (STD) of 5.8 cm, whereas the accuracy of EGM2008 in the high mountainous area is 1.4 cm better than that of XGM2016, which may be attributed to its low resolution. The modification parameters between the two gravity field models showed that the coefficient error of XGM2016 model was lower than that of EGM2008 for the same degree of expansion. The different modification limits and integral radii may produce a centimeter level difference in global mean square error, while the influence of the truncation error caused by the integral was at the millimeter level. The terrestrial gravity data error accounted for the majority of the global mean square error. The optimal least squares modification obtained the minimum global mean square error, and the global mean square error calculated based on XGM2016 model was reduced by about 1~3 cm compared with EGM2008. The refined geoid based on the two gravity field models indicated that both KTH and RCR method can effectively improve the STD of the geoid model from about six to three centimeters. The refined accuracy of EGM2008 model using RCR and KTH methods is slightly better than that of XGM2016 model in the plain and high mountain areas after seven-parameter fitting. EGM2008 based on the KTH method was the most precise at ± 2.0 cm in the plain area and ± 2.4 cm in the mountainous area. Generally, for the refined geoid based on the two Earth gravity models, KTH produced results similar to RCR in the plain area, and had relatively better performance for the mountainous area where terrestrial gravity data is sparse and unevenly distributed.

Watershed-Based Superpixels With Global and Local Boundary Marching

Superpixels are widely used in computer vision applications, as they conserve the running costs of subsequent processing while preserving the original performance. In most of the existing algorithms, the boundary adherence and the compactness of superpixels are necessarily inter-inhibitive because the color/gradient information is balanced against the position constraints, and the set criteria define all pixels indiscriminately. In this paper, we present a two-phase superpixel segmentation method based on the watershed transformation. After designing a new approach for calculating the flooding priority, we propose a new strategy with two distinct criteria for global and local refinement of the boundary pixels. These criteria reduce the compromise between the boundary adherence and compactness. Unlike the indiscriminate standards, our method applies different treatments to pixels in different environments, preserving the color homogeneity in content-rich areas while improving the regularity of the superpixels in content-plain regions. The superior accuracy and computing time of our proposed method are verified in comparison experiments with several state-of-the-art methods.