Recursive N-Dimensional Image Thinning Algorithm And Its Applications

In the field of digital image processing,how to extract the features of image shape information is an important problem.By thinning or skeletonizing the target image,we can simplify its geometric features and analyze the topological structure information.Therefore,the algorithms and applications of image thinning have attracted the attention of domestic and foreign experts and scholars.For an ND target image,the reported mainstream thinning algorithms generally assume that the foreground elements are 3~N-1 connected and the background elements are 2N connected.Under such assumption,with the increase of N,the complexity and operation time of the thinning algorithm will increase dramatically.So the reported classical thinning algorithms are mainly focused on 2D images.Only a few papers have reported the thinning algorithm of 4D images,and the detailed algorithm of 5D and HD images has not been reported.In this paper,a 4-connected Ordered Subfield Algorithm(OSA-4)of2D thinning is proposed,which assumes that the foreground elements are 4-connected and the background elements are 8-connected.On the basis of OSA-4,the complexity of the algorithm was reduced by using the recursive method,so that the complete program of ND thinning algorithm was realized for the first time,which fills up the research gap in the field of HD thinning.The main research work and innovations of this article are as follows:(1)OSA-4 was proposed and completed:the Euclidean distances from all foreground elements to their nearest boundaries are calculated,and then divide foreground elements into different groups according to the ascending order of distance values;in the same group,each foreground element abtained a certain score decided by its neighborhood and location characteristics,and then assigned to specified subfield according to the score;the thinning process first deals with the foreground elements of the low score subfields,and then those of the high score subfileds;in each subfield,parallel processing is used to ensure the deletion or reservation of foreground elements are independent to the internal processing order.After carefully analyzing the 4-connected neighborhood relationship and designing the score lookup table,the new algorithm has the advantages of keeping the topological structure unchanged,unit width,rotation insensitivity and noise insensitivity.(2)We further develop the OSA-2N-ND algorithm with 2N connected foreground elements,and tries to realize any dimensionl thinning by recursion.The key technologies of keeping the topological structure unchanged in the thinning process of dimension reduction were found and proved strictly,which leads to the recursive ND thinning algorithm and program are realized perfectly.Compared with other reported HD thinning algorithms,the OSA-2N-ND algorithm has no breakpoint,does not change the number of closed holes,conforms to the characteristics of human vision,and has a relatively fast operation speed.It can also be used in combination with other reported thinning algorithms to obtain 3~N-1 connected thinning results that are more in line with visual perception.(3)In this paper,the new algorithm is applied to the feature extraction and number statistics of shrimp image to realize a method of shrimp counting based on image processing.Firstly,identify the stomach area of shrimp fry by image processing technology,then used OSA-4 to extract the central axis of shrimp seedling,and calculate its length,node number and other characteristic parameters to obtain the counting result.The experimental results show that the thinning results of the new algorithm have the characteristics of strict single pixel width,and its endpoints,intersections and branches accord with the actual characteristics.The number of shrimps calculated by the algorithm is close to the actual value,and the error rate is stable below 2%,which can greatly improve the working efficiency of shrimp counting.It can be used as an effective processing method to count the number of linear targets and has good application value.(4)The application of the new algorithm in the diagnosis of new coronavirus using CT images is realized.Firstly,3D thinning of blood vessels and diseased tissues is carried out by using OSA-6-3D algorithm.And then,the voxel density of the foreground elements in the 3D images of the thinning results is counted.At last,the location of the diseased tissue is identified.The test results show that this method has two advantages.On the one hand,it is very convenient to identify the lesion area from3D perspective and locate it quickly because of the skeleton processing.On the other hand,we can analyze the characteristics and evolution of the lesion area comprehensively with the help of MATLAB software.Therefore,the new algorithm is expected to be widely used in the field of medical imaging.(5)The application of the new algorithm in the path finding of HD maze is realized.The new algorithm is used to thin the HD labyrinth image,it can make a large number of complex mesh paths are reduced to straight-line path.With the help of zooming and transforming the 3D perspective,the route of HD maze image can be found by the human eye.After using the new algorithm to simplify the path,not only the visibility can be greatly improved,but also the operation speed of the automatic path finding algorithm can be effectively accelerated.Therefore,the new algorithm can helps people better understand the HD space-time,and can be applied to various virtual reality occasions.It is verified that the proposed thinning algorithm can effectively preserve the topological structures of any dimensional target images,and extract the skeletons that conforms to the characteristics of human vision.The results show that the new algorithm has a wide range of application prospects.