Study On Method And Application Of Image Processing Based On Floating Point Coordinate System

In real life,image can convey more real and intuitive information than other ways in human perception,so target recognition using images is one of the hotteat topics in information intelligent system.From the original model matching to the present three-dimensional image recognition,people have been diligently looking for a thorough solution to the model of human cognition both from the hardware and the algorithm.“High dimension of the common goal recognition modeling strategy” is a new idea to solve the problem of image recognition technology in the field of image recognition technology.In view of the current image recognition process cannot completely overcome the effect of the pose,expression and illumination,this thesis adopted a image coordinate system named floating point coordinate system which is independent of the origin of the image coordinate.Firstly floating point coordinate system is consistent with the "HDG" file of high dimensional space geometry,and the data format of the bionic pattern recognition is standardized.The most important aspect is that it fully excavated the image file information,and released the file data from the neatly arranged system,and then image data can not only be combined in accordance with the image content freely,but also maked the image deformation and connection becomes more intuitive,simple and vivid.The research work and innovation of this thesis mainly include:1.The file format of floating point coordinate system is formed.Referring to the common image file format and combining with the basic format of high dimensional geometry data,this thesis formed an image data storage format used in the field of image file processing.Firstly the information of each pixel point extended from "R,G,B" to "AX,AY,R,G,B ".Secondly,in order to reflect the three-dimensional nature of the image,the image content can be layered,and each layer can be combined and connected through connecting points.Different from the existing image file format in which the data form is based on the pixel value,the file format of the floating point coordinate system is based on both of the position information of the pixel and the pixel value information.2.Grid is a commonly model used in image processing.Based on the floating point coordinate system,this paper formed a new triangular mesh model.In each triangle mesh,three vertices are used as the constraint vertex of the coordinate system,and then the remaining points can be determined by two coordinate values.This thesis derived the conversion mathematical models of triangular coordinates and descartes coordinate system,and demonstrated a series of properties of triangular coordinates: affine invariance,continuity,decomposability,linear deformation and rotation,zoom,shift invariance.These properties plays an important role in the field of image processing used the triangle coordinate system.Floating point coordinates and the triangular coordinates will release the describe methods of image from cumbersome and inflexible pixels,and turn to the image content based grid coordinate system,which realized the separation of structure and texture,and then the image processing is more flexible.3.based on the linear non-deformation and rotation,scaling and translation invariance of the triangular coordinate system,this thesis presents an image rectification normalization algorithm based on triangle coordinates.Firstly the normalized algorithm is given in triangular coordinates system;and then experiments are presented on face image pose,expression,zoom correction and normalization in face image processing.4.By using the triangle coordinate deformation continuity properties,this thesis proposes a new image interpolation based on triangular coordinate system,the intermediate image obtained from fewer study samples refined learning samples,so as to improve the effective coverage of biomimetic pattern recognition algorithm.Finally the results proved the algorithm is effective.5.An expression vector coordinate system is proposed.In the triangle coordinate system,.Firstly,the feature points of the image are analyzed in the triangle coordinate system,and the displacement vectors of the seven expressions are obtained in the high dimension space.We proved the direction of the same expression by experiments,and verified the validity of the expression vector coordinate system.Finally,the facial expression recognition used the expression vector coordinate system,verified the validity of our method.