Considering the fact that the smallest eigenvalue of covariance matrix of the raw patches extracted from noise images is significantly correlated with noise level,this paper proposed a fast algorithm that directly uses a pretrained nonlinear mapping model based on the polynomial regression to map (rectify) the smallest eigenvalue to the final estimate.Firstly,some representative natural images without distortion are selected as training set.Then,the training sample library is formed,and the training set images are corrupted with the different noise levels.Based on this,raw patches are extracted for each noisy image,and the smallest eigenvalue of covariance matrix of the raw patches is gotten by PCA transformation.Finally,a nonlinear mapping model between the smallest eigenvalue and the noise level are obtained based on polynomial regression technique.Extensive experiments show that the proposed algorithm works well for a wide range of noise levels and has outstanding performance at low levels in particular compared with the existing algorithms,showing a good compromise between speed and accuracy in general.

In the rail transit of the structured scene,due to the movement of camera,the objects in the image captured by the on-board camera will spread around the center of this image,which is called FOE (Focus of Expansion).In view of the current technology based on FOE,which is sensitive to noise and has a large amount of computation,it can not accurately calculate the FOE in the railway scene.This paper presented a method for estimating the FOE of railway video sequences.This method uses the Pyramid optical flow method to track and coarsely match the detected Harris corner points,and makes accurately matching with RANSAC algorithm based on the computation of fundamental matrix.Then the epipolar lines are extracted in the image,and the FOE is obtained at last.The experimental results show that the FOE error of this algorithm is smaller than that of the Hough line,and the proposed algorithm is suitable for real-time application.

Large cancer genome projects such as The Cancer Genome Atlas(TCGA) and International Cancer Genome Consortium(ICGC) have produced big amount of data collected from patients with different cancer types.The identification of mutated genes causing cancer is a significant challenge.Genovariation in cancer cells can be divided into two types:functional driver mutation and random passenger mutation.Identifcation of driver genes is benefit to understand the pathogenesis and progression of cancer,as well as research cancer drug and targeted therapy,and it is an essential problem in the field of bioinformatics.This paper proposed a driver pathway identification algorithm based on mutated gene networks for cancer(GNDP).In GNDP,a nonoverlap balance metric is defined to measure the possibility of two genes lying in the same driver pathway.To reduce the complexity of the constructed mutually exclusive gene networks,the nonoverlap balance metric,the exclusivity and the coverage of a gene pair are computed first,and then the edges with low nonoverlap balance metric,low exclusivity and low coverage are deleted.Then,all maximal cliques which might be potential driver pathways are found out.After that,the weight of each clique is assigned as the product of its exclusive degree and coverage degree and then every node of a clique will be checked to judge whether is’s deletion might obtain a larger weight.At last,the maximal weight cliques are obtained in mutually exclusive gene networks as the final driver pathways.This paper compared GNDP algorithm with classical algorithm Dendrix and Multi-Dendrix on both simulated data sets and somatic mutation data sets.The results show that GNDP can detect all artificial pathways in simulated data.For Lung adenocarcinoma and Glioblastoma data,GNDP shows higher efficiency and accuracy than the comparison algorithms.In addition,GNDP does not need any prior knowledge and does not need to set the number of genes in driver pathways in advance.

For the purpose of the detection of bad postures and analysis of people’s sitting habit,a sitting posture detection system based on depth sensor was designed.The system first uses the Astra3D sensor to obtain the depth image of human body’s sitting posture and designs the fast and effective foreground extraction method based on the thre-shold segmentation method.The sitting foreground segmentation images are projected into three Cartesian planes respectively and three projection maps are obtained.The background removal,interpolation scaling and normalization are performed for each projection map to obtain the projection features.After the PCA dimensionality,the projection features and the pyramid HOG feature of the front view form the final sitting posture feature vector.Then,random forest is used toclassify and identify 14 kinds of sitting posture.In the experiment,the sitting posture depth image database of 20 people is used for uniform testing and cross testing.The test results show that the method of sitting posture recognition has good recognition rate and recognition speed,and it is superior to the existing method in the type of sitting posture and recognition.Finally,the method was implemented on the Android platform and the application software of the sitting posture detection system was designed to realize the effective detection of sitting posture and timely reminder for the bad sitting posture.

Accurate lung tumor segmentation is critical to the development of radiotherapy and surgical procedures.This paper proposed a new multimodal lung tumor image segmentation method by combining the advantages and disadvantages of PET and CT to solve the weakness of single-mode image segmentation,such as the unsatisfied segmentation accuracy.Firstly,the initial contour is obtained by the pre-segmentation of PET image through using region growing and mathematical morphology.The initial contour can be used to automatically obtain the seed points required for random walk of PET and CT images,at the same time,it can be also used as a constraint in the random walk of CT image to solve the shortcoming that the tumor area is not obvious if the CT image has not been enhanced.For the reason that CT provides essential details on anatomic structures,the anatomic structures of CT can be used to improve the weight of random walk on PET images.Finally,the similarity matrices obtained by random walk on PET and CT image are weighted to obtain an identical result on PET and CT images.Clinical PET-CT image segmentation of lung tumorshows that the proposed method has better performance than other traditional image segmentation methods.

The pros and cons of fuzzy static image target reproduction method directly affect the final effect of fuzzy static image processing and the accuracy of target recognition.At present,the fuzzy static image target reproduction method is used to estimate the ambient light value of the fuzzy static image target and the transmittance of the fuzzy static image target based on the illumination condition.Then,the fuzzy model is used to restore the fuzzy static image target.Finally,the reversion results of the fuzzy static image target are obtained by reversing the target inversion.There is a problem of low target contrast in this method.In order to improve the contrast and the visual effect,a fuzzy static image target reproduction method based on virtual reality technology was proposed.Firstly,fuzzy realistic image and optical imaging principle are used to collect and classify fuzzy static image objects.Then,the gradation adjustment function is used to deal with the brightness channel of the fuzzy static image object,and the global mapping is carried out.Finally,the target details are processed accordingly,the visibility of the target details is maintained,and the target reproduction is completed.The experimental results show that the proposed method has better visual effects and more obvious details.

Traditional Camshift algorithm tracks target characterized by the color histogram,with strong robustness for rigid target tracking.However,the tracking effects and accuracy may not be ideal when they are interfered by the objects with similar color.In view of this,a Camshift tracking algorithm of multi-feature fusion was proposed.Firstly,by extracting and processing the color feature,edge feature as well as the spatial information of target,the color space histogram and spatial edge orientation histogram are obtained.Then the central locations of target matching are obtained respectively under the framework of Camshift algorithm and the weight coefficient is obtained by using the similarity vectors of each image.The optimal central location is obtained through the method of adaptive weighted fusion.The experimental results show that compared with the traditional Camshift target tracking and the improved Meanshift algorithm based on complex feature fusion,the proposed method can effectively overcome the problems of color interference and overlapping occlusion that hamper tracking effects.It can avoid the target loss in the process of tracking which breaks throughthe limitations of traditional methods.

Aiming at the problem that complicated images are interfered by background,an image segmentation method based on saliency and pulse coupled neural network (SPCNN) was proposed.Firstly,with the saliency filtering algorithm and the method of maximum between-class variance,the saliency map and the object image are obtained.Based on this,the interference which comes from the background for the initial seed point selection is eliminated.Secondly,according to saliency values in saliency map,the most saliency region is captured and the initial seed points are produced.Finally,the operations of object image segmentation are achieved via the improved RG-PCNN model.The experimental segmentation results of the gray natural images are obtained from the Berkeley segmentation dataset and ground truth database.There are three evaluating indicators:consistency coefficient(CC),similarity coefficient(SC) and integrate coefficient(IC).The experiment results show that the proposed model has better performance in terms of CC,SC and IC comparing with pulse coupled neural network (PCNN),region growing model (RG) and SPCNN.

There are mainly three challenges for sparse coding in the process of image denoising,including incomplete image denoising,the noise residue,and the lack of protection of image edges and detailed characteristics.This paper proposed a dictionary learning image denoising algorithm combining the second generation Bandelet transformation block method to achieve better removal of noise.With the second generation Bandelet transformation,the sparse representation of images can be automatically obtained to accurately estimate the image information according to the regularity of the image geometry manifold.The K-singular value decomposition (K-SVD) algorithm is used to learn the dictionary under the moderate Gaussian white noise variance.Moreover,it utilizes the quadtree segmentation to adaptively predict the noise images and segment images into blocks.Experimental results show that the proposed method can effectively preserve the edge features of image and the fine structure of image while removing the noise.Since it employs the second generation Bandelet transformation for segmentation,the algorithm structure is well optimized and the operational efficiency is also improved.

In order to improve the performance of the practical pavement crack detection under complex background noise,an improved pavement crack detection algorithm based on sparse representation and multi-feature fusion was proposed.Firstly,this algorithm takes image sub-block as unit,and extracts statistics,texture and shape features which are effective for crack re-cognition.Then,the sparse representation classification method is adopted to realize sub-block re-cognition under each feature matrix separately,and a comprehensive recognition classifier for sub-block detection is designed by fusing the recognition results under different features.Finally,on the detected sub-block,a pixel-level detection method based on visual saliency is used to extract crack details accurately.The experiment results on highway pavement datasets show that the proposed algorithm can effectively improve the accuracy of pavement crack detection and has good robustness.