Study On The Chaotic Attractor Of Image

Currently, a variety of image features are used for image recognition, image tracking,image storage, in the image research field. These characteristics including geometry,algebra, conversion, statistical and neural networks features, which all have their owndistinctive trait, are suitable for the respective space. This paper presents and analyzes theimage features based on iterative, essentially different from the existing features, designscorrespondingly algorithms and experimental, analyzes the experiment results andverifies the feasibility of method. The main research contents are as follows:1. Because the image is a curved surface, so the first study gives the geometricalconditions of chaotic dynamic system required a binary function. According to thefindings of surface iteration, the image and the auxiliary surface constitutetwo-dimensional dynamical system together, in the discrete and continuous case, it isfound that can produce approximately two-dimensional chaotic attractor, which has acertain degree of stability and varies with the change of image;2. Study on the relationship between the change of image and the change oftwo-dimensional chaotic attractor and the variation the changes characteristics of chaoticattractors from the changes of auxiliary function coefficients, theoretical analysis andexperimental verification two-dimensional chaotic attractor as the image features. Then,make this attractor as features use for face recognition, through the observation andanalysis attractor bitmap and by calculation of the correlation coefficient are compared,itverifies that the trend of samples attractor in same set is similar, the attractor graphgenerated by different target image is very different. Correlation coefficient of attractorgraph in same samples set is larger, that of different target image is small; The resultsalso show that the method in algorithm design, computational speed, recognitionaccuracy and robustness have a very good effect, and confirm the two-dimensionalchaotic attractor as image characteristic by the experiments, and for image recognitionconclusion;3. Paper puts forward an algorithm based on three-dimensional function anditerative sequences of image, find sequences of image generate a three-dimensional attractor, make face, handwritten Chinese characters, leaves as experimental subjects,through the observation and analysis of three-dimensional attractor bitmap, and bycalculating the correlation coefficients are compared, we find the three-dimensionalchaotic attractor could be used as the image feature, and for image recognition.4. This paper proposes a method to construct the cosine basis function intwo-dimensional chaotic attractors, and generates the two-dimensional chaotic attractorwith face images and handwritten Chinese character. The experiments found that thecosine basic function of the discretion cosine transform can be used to identify the imageand the chaotic attractor of the image generation and replace the sine function as theauxiliary function.In summary, this paper employs MATLAB tools, the two-dimensional andthree-dimensional chaotic attractors as the image feature make some recognition of theimage, the experimental results show that the method is feasible and has higher accuracyand robustness.