Resize image by shrinking using maximum pixels? #4606
-
|
Is there a strategy by which I can resize images by shrinking them using the maximum pixels? The images are masks and I'd like to retain all 1's over 0's when resized. For However I can't find a similar means with |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 7 replies
-
|
Hello @larsmaxfield, You could run a rank filter before resizing with factor = 5
mask = max.rank(factor, factor, factor * factor - 1).(1.0 / factor, kernel="nearest")That'd do what you want for int factors at least. Not too quick though! |
Beta Was this translation helpful? Give feedback.
-
|
Oh dear, that sounds bad. I'll have a look. |
Beta Was this translation helpful? Give feedback.
-
|
Yes, there's a bug, I made an issue: #4609 The rank filter uses histograms for fast rank filtering with large windows, and this seems to have a serious bug. As a workaround, you can disable this path by eg. changing these lines: https://github.com/libvips/libvips/blob/master/libvips/morphology/rank.c#L454-L462 To perhaps: for (y = 0; y < r->height; y++) {
if (rank->index == 0)
SWITCH(LOOP_MIN)
else if (rank->index == rank->n - 1)
SWITCH(LOOP_MAX)
else
SWITCH(LOOP_SELECT)
} |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for checking that and the workaround. |
Beta Was this translation helpful? Give feedback.
-
|
It was a silly off by one error :( I made a PR and credited you. This fix should be in 8.17.2. #4610 |
Beta Was this translation helpful? Give feedback.
-
|
Thank you for the fix and for the credit even though you did the actual work. A silly mistake in this library is like finding a speck of dust in the glove box of a brand new Ferrari. |
Beta Was this translation helpful? Give feedback.
Hello @larsmaxfield,
You could run a rank filter before resizing with
nearest. In python:That'd do what you want for int factors at least. Not too quick though!