Research And Implementation On Content-based Pathology Image Retrieval

With the rapid development of medical and computer image processing technology,automatic retrieval of pathology image diagnosis is a current research focus in the cross field of computer image processing technology and medical.Using computer image processing technology to complete the identification,retrieval and pathological diagnosis,it purpose is to construct a fast,high right rate pathological image retrieval system and help pathology experts to completing diagnosis.In this paper,human liver tissue pathology image are used as the object,splitting the whole process of retrieval into two stages:firstly,preliminary classification process are used to the pathology image database,then,retrieval based on the images categories which are pre-classified in the steps one.During the process of pre-classification,thesis propose a new method which can deal with multi-classification issues named PSO-ELM,using for pathology image classification before retrieval.It can greatly improve the efficiency and accuracy of retrieval.In addition,if chose a single feature amount for the content-based pathology image retrieval,we will lose a lot of information.The color features and texture features both are very important image feature amount of colorized pathological images,the paper used color histogram-based for color features and LBP-based for texture features.Finally,combine two features using similarity search,presenting a synthesis based on color-texture features for pathology image retrieval.Based on the above algorithm,this paper shows a pathology image retrieval system.This system divided into pathology image input module,feature extraction module,preliminary classification module,retrieval and output module.The experiments proved that the system can retrieval pathology images greatly and it run normally.