文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 自然科学 > Research and Implement the Algorithm of Image Retrieval based on the Multi-features
侵权投诉

Research and Implement the Algorithm of Image Retrieval based on the Multi-features

The paper introduced the algorithm of image retrieval witch combination the features of color, texture and shape and so on.Firstly,the internal and exterior normalization of multi-features are studied and analyzed,and introduced the normalization algorithm in detail. Secondly, the retrieval methods of the multi-features are showed.Lastly,adopted the Euclid distance metric.

2010 International Conference on Challenges in Environmental Science and Computer Engineering

Research and Implement the Algorithm of Image Retrieval based on the Multi-features

GUO Xiaojuan, ZHU Yanli, WANG Yufen and CHENG Jifu

School of Information Engineering Henan Institute of Science and Technology Xinxiang, Henan Province 453003, China

gxj1315@126.com

Abstract—The paper introduced the algorithm of image features feature vector to be the same in the similar distance retrieval witch combination the features of color, texture and computation. When uses Euclidean space is to carry shape and so on. Firstly, the internal and exterior synthesis similar measure, the different feature vector will normalization of multi-features are studied and analyzed, and be deal with exterior normalization. introduced the normalization algorithm in detail. Secondly, the

① Internal normalization retrieval methods of the multi-features are showed. Lastly,

The goal of the process of feature interior adopted the Euclid distance metric, and used the UIUC image

normalization lie in the feature vector various database, the multi-features retrieval algorithm have been

components to have same important [5]. Because the confirmed and compared. From the experimental result, we

know that use the multi-features retrieval method can obtain identical feature vector's different component possibly the good retrieval effect. has the completely different physics significance, they

respective change scope possibly has the very big

Keywords-image retrieval;features extraction;similarity difference, if does not pass through the normalized metric; normalization; multi-features

operation and to calculate directly similarity, will cause the big deviation. Therefore has the necessity

I. INTRODUCTION the feature vector various components normalization

to the consistent scope. In fact, any the image retrieval based on single feature

If a N dimension feature vector is has certain limitation, it cannot the comprehensive descript

the image contain content [1]. If the multi features of image Fi=[F1,F2,...,FN], if use I1,I2,...,IM make in the integration image retrieval system, the effect of

represents in the image, then may record N×M as the image retrieval will obtain enormous enhancement [2].

a matrix, corresponds any image its corresponding But description methods such as color, texture, shape had

feature vector is Fi=[F1,F2,...,FN], then the reflected the image feature from the different angle, how

does organize these characteristics, enables them to defer to matrix kth row is the row vector of M dimension, the user’s request to merge each kind of feature, this is a

records Fk, any vector representation Fk,m, and may

question which needs to study.

use the Gauss normalization method.

II. FEATURE EXTRACTION AND NORMALIZATION This method is one good normalized, its characteristic

is few ultra big or the ultra small element to influence little

A. Feature Extraction to the entire normalization element value after distribution,

For the color, the texture and the shape extract separately and enables the each component of feature vector have the its feature: color moment, co-occurrence matrix and Hu equating importance. The supposition feature component Moment Invariants, its algorithm sees Ref [3]. Under value series conforms to the Gaussian distribution, below the paper detailed introduce normalization calculates the average value m and the standard deviation

kprocessing question of these features vector.

σk, then the feature component normalization sector, the

B. Normalization

concrete method is as follows:

The different feature and the identical feature's different Mcomponent represent different physics significance, the ∑Fk,mvalue scope is also widely divergent, such different feature average value: m=m1

k

has not the commeasurable [4]. Therefore needs normalize Mthe feature before to carrying on the retrieval. The

normalized method divides into the internal normalized standard deviation:

Research and Implement the Algorithm of Image Retrieval based on the Multi-features

σmethod and the exterior normalized. The internal

k

normalization's goal is causes the ponderance of the interior

feature vector is the same when the retrieval; But the feature vector exterior normalization is carries on the normalization to the different feature vector, its goal is causes the synthesis

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email