Feature Extraction And Analysis For Xinjiang Uyghur Medicine Image And The Design Of Retrieval Platform

Objective: To extract Xinjiang Uyghur medicine image feature and analysis the features, and the content-based image retrieval will be proceeded by using the features which were researched. To investigate the classification effect and retrieval effect of the researched features in the Uyghur medicine image retrieval and to find the suitable feature for Uyghur medicine image retrieval. To lay the foundation for content-based medical image retrival system in Xinjiang Uyghur medicine image. Methods: Through the Matlab image processing software, firstly, processing of medicine image, to improve the quality of the image, saves the valid information and deletes unnecessary information. And then extract the feature of the processed image based on the color histogram, color moment and shape. Through the SPSS statistical analysis software, using the image features to analyse the image classification. And, furthermore, using these features for content-based image retrival in Xinjiang Uyghur medicine image. Results: To classify the flower image and leaf image of Xinjiang Uyghur medicine image,the result showed that the correct classification rate are 73% and 77% by using the color histogram feature,and the correct classification rate are 79% and 88% by using the color moment feature, the correct classification rate are 56% and 78% by using the shape feature. To classify the flower image and fruit image of Xinjiang Uyghur medicine image,the result showed that the correct classification rate are 62% and 66% by using the color histogram feature,and the correct classification rate are 77% and 72% by using the color moment feature, the correct classification rate are 57% and 66% by using the shape feature. To classify the leaf image and fruit image of Xinjiang Uyghur medicine image,the result showed that the correct classification rate are 77% and 74% by using the color histogram feature,and the correct classification rate are 89% and 74% by using the color moment feature, the correct classification rate are 70% and 61% by using the shape feature. To classify the flower image and leaf image and fruit image of Xinjiang Uyghur medicine image,the result showed that the correct classification rate are 61% and 74% and 60% by using the color histogram feature,and the correct classification rate are 64% and 81% and 61% by using the color moment feature, the correct classification rate are 59% and 66% and 58% by using the shape feature. To retrival the flower image and leaf image and fruit image of Xinjiang Uyghur medicine image,the result showed that the average precision rate are 68% and 58.5% and 65% for the flower image and leaf image and fruit image by using the color histogram feature,and the average precision rate are 67.5% and 71% and 77% for the flower image and leaf image and fruit image by using the color moment feature, and the average precision rate are 57% and 70% and 51.5% for the flower image and leaf image and fruit image by using the shape feature.Conclusion: Image color histogram feature color moment feature and shape feature extraction method have been successfully introduced into the analysis of Xinjiang Uyghur medicine image, extracted the image features, analysed the image by discriminant analysis and content-based image retrival analysis. Through the analysis results show that the color moment feature classfy effect is better for the three types of the medicine image classification and meet the classification requirements to some extent. Using these features for content-based image retrival.For the flower image retrieval,the retrieval effect is better when using color histogram feature, selecting the Mahalanobis Distance.For the leaf image retrieval,the retrieval effect is better when using color moment feature, selecting the Euclidean Distance.For the fruit image retrieval,the retrieval effect is better when using color moment feature, selecting the Standardized Euclidean Distance. The results can be further been applied to content-based Xinjiang Uygur medical image retrieval system and has a certain positive value.