Contour Statistics Files


*.pdf

    Relevant papers


images/*.png

    Kodak color images

    Notes:

        Some image files do not contain segmentation data from all subjects.


edges/*.dat

    Edge data

    Notes:

        Row 1 contains edge extractor metadata.

        The remaining rows contain records of the form:

            x_coordinate y_coordinate orientation response group_number

        Coordinates are in pixels.

        Orientation is in radians and goes counter-clockwise from 0 to 2pi.

        An edge with an orientation of 0 radians is a horizontal edge whose
        darker pixels lie above the edge and whose lighter pixels lie below the
        edge.


kodak_counts_16_8_72_72.txt

    Edge data histogram

    Notes:

        The histogram contains 16 distance bins, 72 direction bins, and 72
        orientation difference bins.  Distance bins have 8 pixels per bin.
        Direction and orientation difference bins have 2*pi/72 radians per bin.

        Column 1 contains counts of edge pairs on the same contour.

        Column 2 contains counts of edge pairs from different contours.

        Row 1 contains counts for histogram bin (0, 0, 0).  Subsequent histogram
        indices are found by first incrementing the orientation difference bin
        index, and then incrementing the direction bin index, and then
        incrementing the distance bin index.


code/*

    Source code


kodak_table.txt

    Edge data table

    Notes:

        Each row of this file contains the following values:

            d phi thd L beta

        where

            d is the distance in pixels
            phi is the direction in degrees
            thd is the orientation difference in degrees
            L is the distance dependent likelihood
            beta is the ratio of priors as a function of distance

        except the first row which contains column labels.

        This file was created by one of the test functions in the code directory.


segmenting/*

    Examples of image segmenting

    Notes:

        Each file contains images in three rows.

        The first row shows subregions of the original images.

        The second row shows the subregions with color information removed and
        edge pixels marked in yellow.

        The third row shows the subregions with edge groups indicated by their
        color.