Research On The Image Content Ecognition Technology Based On Wavelet Moment

Application of the image has been deep into all aspects of people's lives,how to obtain the needed information from the complex images is difficult to complete manually.In order to retrieve images automatically,the key of that is the image content recognition.The content of images include color, texture, shape and other visual feature.For the wavelet analysis has good time-frequency localization properties, it has been widely used in the expression of the image feature.For geometric invariance of the invariant moment,it has been used in pattern recognition, computer vision and image reconstruction.The wavelet moment is a combination of the wavelet analysis and the invariant moment, it has the advantages of the both of them,so it is valuable for the engineering applications.The research work focuses on the wavelet moment and content-based image recognition.A method of image recognition based on wavelet moment is proposed. The principal research work are listed as follows:1. The typical system of image retrieve are introduced.Their characteristics and application areas are analyzed.The advantages and disadvantages of the color,texture,shape feature of the image are compared.The methods of how to extract color,texture,shape feature of the image are analyzed and the typical methods are investigated.2.To solve the lack of translation, rotation and scaling invariance,when wavelet analysis used in content-based image reconstruction,the wavelet moment that combined of the wavelet analysis and the invariant moment is introduced.The wavelet moment is used to extract the feature of images.The software based on the wavelet moments is designed and the part of image feature extraction is completed.3. Typical images that added noise and rotated are used in the experiment. The image features are extracted by the method of the wavelet moment..Compared with the zernike moment,the wavelet moment features of image have a better rotation invariance and noise immunity.4. The image database used in the experiment is set up.The wavelet moment features of image are extracted.Euclidean distance is used to measure the similarity of different images. The experiment proved that the image content recognition based on wavelet moment is feasible and advanced. In the case of noise, the method of the wavelet moment provides higher recall rate and precision rate than the zernike moment.