Research on tumor image retrieval system based on relevancy discriminant by color feature
Hongbin Zhang · Jisuanji gongcheng yu sheji · 2012
The working flow of a medical image retrieval system is put forward firstly,then some image features in a tumor image are analyzed and color features are chosen as the basic design principle to make correct image retrieval.Some pretreatment steps must is done firstly such as color space selection,color space transformation,color quantification and color features selection.With the aid of pretreatment,three discriminant algorithms such as Euclidean Distance,Improved Euclidian Distance and Cross Color Histogram Statistics are in turn designed and realized to calculate the relevancy distance.Finally,the performances of each algorithm are gotten and some conclusions are also deduced.The time processing of each image is no more than 0.432s.While many recall rates are gotten that the recall rate of the normal images is superior to any other kinds of images.Retrieval accuracy of normal image is higher than that of cancer image.