The Study Of Image Segmentation Application Based On The Forward Cloud Model

Image segmentation is the key technology of image from the image processing to imageanalysis,which is paid high attention to and researched by many scholars. Now,there still notbe a way among numerous Image segmentation methods to be able to solve all the problems.Following the deterministic approach to image segmentation is one of the many key issues ofthe problem of image segmentation. Traditional solution,such as fuzzy theory or probabilitytheory can not really solve this problem. Cloud theory based on fuzzy theory and probabilitytheory can well solve the problem of the "hard segmentation of image segmentation. So itdeserved in-depth study and exploration from many scholars in this field.Initially the cloud theory is applied to spatial data mining. After2006, the cloud theory isused in image processing, and becomes a new method of image segmentation. The coremethod of the Cloud theory is cloud transform. The three basic characteristics of cloudtransform are expectations, entropy, super-entropy. The key problem during the image datacloud processing is the acquisition of this three basic characteristics what are critical todetermine whether the subsequent processing of the image data could get good segmentationthreshold of image processing, so the cloud model is very pivotal. There are some questionsof the cloud model in image processing build process: First, there are inevitable errors duringobtaining entropy in the cloud model, which sometimes can cause a cloud model buildingfially; second, the single cloud core extractint of the cloud model during image processingmay lead to the lost of useful data or the undesirable choose of cloud cores. Besides,it isattempted to combine a one-dimensional cloud model with NBS distance to segment colorimage.According to the inherent problems existing in the process of cloud model building andimage data becoming cloud of cloud model, this article proposes the following improvements:(1).Under in-depth understanding of technically of the cloud theory, on the analysis theerror of built of the cloud model is easy to produce that is cloud-entropy analysis, new methodfor getting cloud entropy is proposed understand the based cloud theory of "3En".The newmethod of entropy is simple,and the acquisition of new entropy can successfully avoidbuilding cloud model. Bsides, it can get entropy of smaller erros through fine-tining entropy according to actual situation.(2).Through the study of traditional entropy method for getting to build cloud modelused to process the image information based on the traditional cloud model multi-thresholdsegmentation method is applied to the image, and demonstrate traditional cloud model imagemulti-threshold segmentation which multi-threshold segmentation is feasible. The new way ofbuild cloud model is applied to graphics multi-threshold segmentation, and goodsegmentation results is obtained by simulation.(3).In-depth understanding of the cloud theory image segmentation methods and steps,according to the cloud model building rules to determine all can locate the position of thecloud cores. Marked the current image data to meet the cloud of nuclear extract position andits cloud processing. Sequential know all data cloud after, and the transportation and treatmentof the segmentation threshold to complete the image segmentation. Experiments show thatthis method is applied to a multi-threshold image segmentation which can obtain goodsegmentation results.(4).Preliminary understanding of the principle of the color image and imagequantization,The cloud model is achieved by a combination of one-dimensional cloud modeland Euclidean distance method,Cloud merger will meet the conditions of the cloud model.Finally, with great discrimination method for the guidelines, the division of the complete colorimage color image pixels divided.This method can complete a good color imagesegmentation, is a new solution for the color image information uncertainty.