Protocol:
                                                                           
All registered participants is required to submit two executables (one for text line detection and one for word detection)  in the form of Win32 console application:

TextLineDetect arg1 arg2

WordDetect arg1 arg2

arg1: a B/W TIF image having CCITT Fax4 compression (or any B/W image format if TIF/CCITT Fax4 is not possible for the participants)
arg2: The output raw image files must have the same format as the Ground Truth Data files which is the following:
For a binay image:

which has segmentation results:

the output raw image file must be:

In the raw image file, an integer is used for every pixel value and  0s (or <0) correspond to the background and all other IDs (>0) define the resulting different segmentation regions. An example of C++ code to read the output raw image files follows:


    int Ix; //Image Width (x=0...Ix-1)

    int Iy; //Image Height (y=0...Iy-1)
    unsigned int *IM_SegmResult; //Pointer to store raw data
    AnsiString SegmResultName; //File that contains the raw data
    FILE *f1;

    IM_SegmResult = (unsigned int *) calloc (Ix*Iy,sizeof(int));
    f1 = fopen(SegmResultName.c_str(),"rb");
    fread(IM_SegmResult,Ix*Iy,sizeof(int),f1);
    fclose(f1);





Site maintained by N. Stamatopoulos, email: nstam<at>iit<dot>demokritos<dot>gr
Last update: 07/01/2013