Brain medical image registration is a fundamental premise for the computer-assisted treatment of brain diseases. The one most important and complex organs human body, it very challenging to perform accurate fast on it. Aiming at problem voxel folding in deformation field low accuracy when facing fine objects, this paper proposed fully convolutional multi-constraint cascaded attention network (MCANet). composed two sub-network cascades performs coarse-to-fine input pairs an iterative manner....

In recent years, deep learning has made successful applications and remarkable achievements in the field of medical image registration, method registration based on become current research hotspot. However, performance convolutional neural networks may not be fully exploited due to neglect spatial relationships between distant locations incomplete updates network parameters. To avoid this phenomenon, MHNet, a multiscale hierarchical deformable for 3D brain MR images, was proposed paper. This...

This paper introduces a new nonrigid registration approach for medical images applying an information theoretic measure based on Arimoto entropy with gradient distributions. A normalized dissimilarity is presented, which employed to the independence between two images. In addition, regularization term integrated into cost function obtain smooth elastic deformation. To take spatial voxels account, distance of distributions constructed. The goal alignment find optimal solution including...

Target detection and tracking of the video is an important part intelligent monitoring system. This paper proposes algorithm which uses matching trajectory to realize target based on Gaussian mixture model. Firstly we use background subtraction detect target. Secondly, feature color information, position shape information modified Hough arithmetic are used find trajectory. At last, get result tracking. xperimental shows method has shorter time good rate in

Key frame extraction is an important component of content-based video retrieval and directly influences on the efficiency retrieval. Nowadays some problems are existed in algorithms key-frame extraction, such as features selected singly, choosing threshold value difficultly so on. This paper proposes a new method which based adaptive detection multi-features. First, two descriptors color histogram, edge histogram adjacent scale wavelet transform used to describe visual content, combine form...

It is important to extract ROI (region of interest) based on vision image processing in intelligent transportation system. A simple and novel approach proposed the RIO scene images. First, interest pixels function designed detect all from image; after that, these are grouped regions according some strategy; Last, selected by region threshold as result extraction RIO. The experiment presented last part paper feasible valid.

In ITS (intelligent transportation system), on-road object detection algorithm is one of the most important research fields and obstacle segmentation a key factor in approaches. this paper, simple fast approach proposed for ITS. Firstly, method applied to detect interest pixels scene images by defined function single strategy reduce redundant computation process gray mean squared window; Secondly, all detected neighbored each other are grouped region; after that, non-interest enclosed region...

According to the low recognition rate of Hu invariant moments in target images, this article proposes a vehicle-logo research algorithm based on modified moments. At first, use template matching locate rough area and edge detection for accurate location. Then, calculate characteristic value vehicle-logo, finally, is recognized according minimum distance The experimental results show that can improve images effectively.

This paper proposed a simple and novel approach for on-road object detection based on vision. Firstly, method is applied to detect the interest pixels of in images by defined function single strategy reduce redundant computation process gray mean squared window; Secondly, all detected neighbored each other are grouped region; after that, non-interest set binarized; The binary could be easily segmented applying contour tracking algorithms. Lastly, several captured from ITS or ALV system used...

Shot boundary detection is the first step of analysis video contents. In this paper, it proposes a new algorithm shots edge based on product adjacent scale wavelet transform. At first, each frame image in shot carried transform, and then gets modulus direction transform plot every image. Secondly, we calculate absolute distance among consecutive frames modulus, get discrete value between two frames. Finally use adaptive threshold segmentation to detect edges.

This paper studies the problem of stability analysis for singular systems with time-varying delay. By developing a delay decomposition approach, information delayed plant states can be taken into full consideration, and new delay-dependent sufficient criteria are obtained in terms linear matrix inequalities (LMIs). The merits proposed results lie their less conservatism which is realized by choosing different Lyapunov matrices decomposed integral intervals taking consideration. A numerical...

With the rapid development of online social networks recently, more and users have participated in network activities rich relationships are formed accordingly. These provide a data source research basis for in-depth study on recommender systems (RSs), while also promoting RSs based networks. To solve problems cold start sparsity RSs, many recommendation algorithms constantly being proposed. Motivated by availability connections today’s large number techniques been proposed achieving good...

Twin Parameter-margin support vector machine (TPMSVM) is a recent very powerful binary classifier. To improve its sparsity, linear sparse TPMSVM (Lin-STPMSVM) proposed in this paper. In the primal problem, vectors defining hyperplane are replaced with their expression terms of dual variables as derived from Karush Khun Tucker (KKT) conditions. Then new problems directly optimized, thus ensuring sparsity solutions. Numerical experiments show that solution obtained by model more without...

Fine-grained image classification is a popular research topic in computer vision and pattern recognition, where the goal to recognize classify subclasses of objects images at fine-grained level. In recent years, Transformer's self-attention mechanism has been increasingly introduced into tasks due its ability naturally focus on most discriminative regions object. this paper, new Convolutionally Enhanced Feature Fusion Visual Transformer method proposed based by introducing convolutional...

A new fast area-based stereo algorithm suitable to large disparity pair matching and real time applications is proposed in this paper. Firstly, the approach estimates global of pairs gets two maximum correlation areas left right images estimation level. Then, based on area used match disparity, computation optimization strategy applied speed algorithm, sum computed result disparity. Finally, paper provides compared experiments. The results indicate that combining with calculation scheme...

In ALV system based on stereo vision, the obstacle detection is one of most important problems. this paper, an algorithm proposed moment invariant to detect in system. Firstly, a simple method applied binarize images by defined binarization function; Secondly, binarized are segmented using outer contour tracing and regions left right detected respectively; Lastly, matched between image, successful likely obstacle. Stereo pairs captured from used test algorithm, result indicate that approach...

In ALV system, obstacle detection algorithm is one of the most important research fields and segmentation a key factor in approaches. this paper, simple fast approach proposed for system based on binocular stereo vision. Firstly, method applied to binarize images by defined binarization function; Secondly, order save computation time gray mean pixels squared window, reduce redundant our algorithm; Lastly, avoid affection inner region objects, outer contour tracing get likely binarized...

Tchebichef discrete orthogonal moment can reconstruct image, but there are errors between the reconstructed image and original image. In fact, error information includes some valuable information. A new stereo matching approach based on reconstruction pair is proposed in this paper. Firstly, moments employed to coordinate space, after that, computed by difference of finally, disparity images space correlation window approach. The result indicates that data be applied matching.

Fine grain image is an important research field of recognition, which can classify objects in more detail. Image feature extraction usually implemented by bilinear network, meet the desired effect extraction, but also discard some detailed information. In this study, a fine-grained classification optimization algorithm that combines target detection method YOLOv7 and multi-feature fusion network proposed to achieve classification. method, used locate target, recognition content extracted...