[Octave-bug-tracker] [bug #50188] image package: poly2mask does incompatible rounding

[Hartmut]

URL:
  <http://savannah.gnu.org/bugs/?50188>

                 Summary: image package: poly2mask does incompatible rounding
                 Project: GNU Octave
            Submitted by: hardy
            Submitted on: Mon 30 Jan 2017 08:07:25 PM GMT
                Category: Octave Forge Package
                Severity: 3 - Normal
                Priority: 5 - Normal
              Item Group: Matlab Compatibility
                  Status: None
             Assigned to: None
         Originator Name: 
        Originator Email: 
             Open/Closed: Open
         Discussion Lock: Any
                 Release: 4.2.0
        Operating System: Any

    _______________________________________________________

Details:

This happens with the current image-2.6.1 package and Octave 4.2.0.

The current Octave implementation of poly2mask.m rounds its polygon input
values to the nearest integer (and correctly says so in its documentation).

In contrast, the Matlab implementation of poly2mask.m also uses fractional
input values, and does some 5-by-5 subsampling of pixels. See the Matlab
documentation:
https://de.mathworks.com/help/images/ref/poly2mask.html#f6-465457

As a result the output values between the Octave and the Matlab version of
poly2mask are different.

Here is some typical code that shows the difference, and can be used for
implementing an improved version of Octave's poly2mask:


clear;

poly_a = [...
    2   2.5
    1.5 3
    2   3.5
    4   3.5
    4.5 3
    4   2.5
    2   2.5];

mask_a = poly2mask( poly_a(:,1), poly_a(:, 2), 5, 5)

mask_a_matlab = logical([ ...
     0     0     0     0     0
     0     0     0     0     0
     0     1     1     1     0
     0     0     0     0     0
     0     0     0     0     0])

poly_b = [...
    2 2
    9 4
    9 5
    2 5
    2 2];

mask_b = poly2mask( poly_b(:,1), poly_b(:, 2), 6, 11)

mask_b_matlab = logical([ ...
     0     0     0     0     0     0     0     0     0     0     0
     0     0     0     0     0     0     0     0     0     0     0
     0     0     1     1     1     0     0     0     0     0     0
     0     0     1     1     1     1     1     1     0     0     0
     0     0     1     1     1     1     1     1     1     0     0
     0     0     0     0     0     0     0     0     0     0     0])


The current output of Octave is:


mask_a =

  0  0  0  0  0
  0  0  0  0  0
  0  0  0  0  0
  0  0  1  1  0
  0  0  0  0  0

mask_a_matlab =

  0  0  0  0  0
  0  0  0  0  0
  0  1  1  1  0
  0  0  0  0  0
  0  0  0  0  0

mask_b =

  0  0  0  0  0  0  0  0  0  0  0
  0  0  0  0  0  0  0  0  0  0  0
  0  0  1  1  1  1  0  0  0  0  0
  0  0  1  1  1  1  1  1  1  0  0
  0  0  1  1  1  1  1  1  1  0  0
  0  0  0  0  0  0  0  0  0  0  0

mask_b_matlab =

  0  0  0  0  0  0  0  0  0  0  0
  0  0  0  0  0  0  0  0  0  0  0
  0  0  1  1  1  0  0  0  0  0  0
  0  0  1  1  1  1  1  1  0  0  0
  0  0  1  1  1  1  1  1  1  0  0
  0  0  0  0  0  0  0  0  0  0  0


(I hit this bug while implementing a new feature in regionprops.m. I will
probably work around this bug in regionprops, to not introduce more
dependencies of patches. This means that a future fix to this bug could have a
look what's been done in regionprops.m for the 'ConvexImage' calculation.)





    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?50188>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/