Bug 375142 - The similarity calculated for the fuzzy image drop search is imprecise
Summary: The similarity calculated for the fuzzy image drop search is imprecise
Status: RESOLVED FIXED
Alias: None
Product: digikam
Classification: Applications
Component: Searches-Similarity (show other bugs)
Version: 5.4.0
Platform: Compiled Sources Linux
: NOR normal
Target Milestone: ---
Assignee: Digikam Developers
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2017-01-16 15:11 UTC by Mario Frank
Modified: 2017-01-16 15:15 UTC (History)
1 user (show)

See Also:
Latest Commit:
Version Fixed In: 5.5.0


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Mario Frank 2017-01-16 15:11:05 UTC
When dropping an image from outside of the digiKam database into the fuzzy search
while an identical image is present in digiKam, the similarity is not even near 100 % though it should be.

The problem seems to be some magic during loading the image as QImage. This seems to reduce the precision.

This file is opened for reference.
Comment 1 Mario Frank 2017-01-16 15:15:41 UTC
Git commit c97ac972bff7914efd94202ef26eb2d5bfb8ff66 by Mario Frank.
Committed on 16/01/2017 at 15:11.
Pushed by mfrank into branch 'master'.

Improved the exactness of fuzzy image drop search.
Some images which were present in DB and outside had only similarity of 98,13 % though they were identical.
The problem seems to be that QImage does some transformations which reduce the exactness.
Now the dropped external image is loaded as DImg which increases the similarity of the specific pictures to 99,41 %.
This is still not perfect but far more precise.
FIXED-IN: 5.5.0

M  +4    -2    NEWS
M  +1    -6    libs/database/haar/haariface.cpp

https://commits.kde.org/digikam/c97ac972bff7914efd94202ef26eb2d5bfb8ff66