A comparison of methods for extracting information from the co-occurrence matrix for subcellular classification

摘要: Abstract In this paper we focus on cell phenotype image classification, a bioimaging problem that is concerned with finding the location of protein expressions within cell. Protein localization becoming increasingly critical in diagnosis and prognosis many diseases. recent years several new approaches for describing given have been proposed. Some most significant developments based binary encodings, such as local patterns phase quantization. reexamine one oldest methods representing an Haralick famously proposed 1979 using co-occurrence matrix calculating set statistics. Few since extract features from matrix. work compare some recently are (CM) to classify images. We investigate correlation among different sets can be extracted CM then determine best way combine these feature optimizing system performance. Moreover, our novel approach state art descriptors optimize validate various types biological microscope images five databases subcellular classification. use training stand-alone support vector machine random subspace machines separate classes each dataset. The Matlab code tested will available at http://www.dei.unipd.it/wdyn/?IDsezione=3314&IDgruppo_pass=124&preview= >.