Bashar-Hozaim/Writer-Identification-using-Textural-Features

This project is about the writer identification for handwritten text in Arabic and English Training_KHATT_Writers.mat, Testing_KHATT_Writers.mat, Training_IAM_Writers.mat, and Testing_IAM_Writers.mat files contain the Arabic KHATT and English IAM handwritten connected components written by 500 scribes per each dataset. the files are already preproceesed and the connected component images are saved as MAT files. Raw datasets are fully accessible and freely available for reseraches purposes. IAM Dataset can be easily downloaded from the following link: fki.tic.heia-fr.ch Khatt dataset is available and can be downloaded using this link: khatt.ideas2serve.net ------------------------------------------------------------------------------------------- src.rar file contains the source code files used to implement the writer identification system. The structure of source code files is as follows: Writer_IdentiFication_System M script is used as the main module of the system. through this script, you can specify which target dataset, how many scribes are involved in the expirement, which feature you need to apply *** Dependencies --> System_Main_Function M script, System_Main_Function M script was used to invoke the module of features extraction, invoke the module of generating the codebook, invoke the module of classification *** Dependencies --> Build_Test_Train_Data_Sheet_1 M script, the description of this script is shown below. --> Calculate_Occurrence_Histogram M script, was used to build a histogram for each reference and questioned document. --> Calculate_Matrices_Similarities and Calculate_Performance M scripts were used in the classification process. Build_Test_Train_Data_Sheet_1 M script was used to extract the features and save them as mat files for further processing. But before features extraction, this module is used to specify the most dominant points, extract batches around the most dominant points, use these batches to build new tex …