Problems with "nearest" procedure in IMROTATE.M

[Journeaux, Ian]

I have be trying to troubleshoot a problem with the imrotate method in Octave.

When I use the "nearest" option, the results of imrotate are not ideal.  I have 
a small test image that I rotate by a few degrees. The histogram of the 
resulting images look ok visually but if you do a histogram of the 8bit 
grayscale values, the Matlab version of imrotate produces a smooth distribution 
that is identical to that of the original image (as expected). The Octave 
version gives an odd looking histogram where many of the bins are 0. It looks 
like there is a significant rounding error.

The table list the histogram distribution for the original image and the  
rotated image via Matlab and Octave versions of imrotate.

Notice that for Octave bin 71 is the sum of bin 71 and 72 in the original image 
and that the Octave bin 72 is the value of the bin 73 in the original image and 
the Octave bin 73 is 0. This pattern repeasts through the histogram.

Bin     Original        Matlab  Octave
64      37      37      38
65      38      38      0
66      38      38      38
67      33      33      65
68      32      32      28
69      28      28      0
70      47      47      47
71      30      30      54
72      24      24      34
73      34      34      0
74      27      27      27
75      32      32      95
76      63      63      80
77      80      80      0
78      190     190     189
79      448     448     1523
80      1087    1087    2422
81      2453    2453    0
82      3922    3923    3868
83      4338    4338    8400
84      4161    4161    3693
85      3741    3741    0
86      2908    2908    2871
87      2383    2383    4145
88      1802    1802    1418
89      1431    1431    0
90      1235    1235    1225
91      1106    1106    2116
92      1028    1028    903
93      911     911     0
94      762     762     759
95      659     659     1177
96      530     530     404
97      410     410     0
98      364     364     364
99      325     325     582
100     259     259     247
101     248     248     0
102     212     212     211
103     192     192     359
104     169     169     152
105     153     153     0
This electronic message contains information from NewPage Corporation or 
subsidiary companies, 
which may be confidential, privileged or otherwise protected from disclosure. 
The information is 
intended to be used solely by the recipient(s) named. If you are not an 
intended recipient, be 
aware that any review, disclosure, copying, distribution or use of this 
transmission or its 
contents is prohibited. If you have received this transmission in error, please 
notify NewPage 
immediately at address@hidden