Welcome to iraf.net Friday, April 26 2024 @ 02:48 PM GMT
Buglog #178
- Tuesday, August 18 1992 @ 05:31 PM GMT
- Contributed by: admin
- Views: 1,036
MODULE: images.fmedian
SYSTEM: V2.9
SYSTEM: V2.9
NUMBER: 178 MODULE: images.fmedian SYSTEM: V2.9 DATE: Tue Aug 18 11:31:49 MST 1992 FROM: davis BUG: The fmedian task can crash with a segmentation violation or a floating operand error if image pixel to integer mapping is turned off (hmin = zmin and hmax = zmax) and the input image contains data outside the range defined by zmin and zmax. The bug is not seen often because for most ccd images the data is between -32768 and 32767. STATUS: The bug is fixed in 2.10.1. The work-around is to either let zmin and zmax default to the image minimum and maximum or set hmin and hmax to numbers that are slightly different from hmin and hmax. NUMBER: 276 MODULE: imexamine SYSTEM: V2.8-V2.10.2 DATE: Tue Jan 31 15:06:41 MST 1995 FROM: valdes BUG: The standard deviation values returned by the 'm' key in IMEXAMINE can be inaccurate when the mean data value is large and the standard deviation is relatively small. This is due to calculation of the standard deviation formula using the difference in the sum of squares and the square of the sum with real precision. The workarounds are to use IMSTATISTICS with an image section or to first subtract something like the mean value from the image before using IMEXAMINE. STATUS: In V2.10.3 and after the calculation is done in double precision and the standard deviation is as accurate as IMSTATISTICS. NUMBER: 450 MODULE: imfilter.fmedian,imfilter.frmedian,imfilter.fmode,imfilter.frmode SYSTEM: V2.11 DATE: Fri Nov 19 14:20:16 MST 1999 FROM: davis BUG: The fast median / mode filtering tasks will overflow producing pathological pixel values if the size of the filtering kernel xfilter * yfilter > 32767 and the number of pixels in 1 histogram bin happens to be > 32767 as may happen in bad pixel regions. The problem was due to the fact that for historical reason the algorithm histogram was stored as a short instead of an int. STATUS: Fixed for 2.11.3. The workaround is to try and keep the filter size < 32767 pixels in size, or to use the much slower median task.
The following comments are owned by whomever posted them. This site is not responsible for what they say.