Comments (9)
lukasalexanderweber
commented on August 21, 2024
Hey,
I also thought about informing the user if not all of the images are stitched.
I think it should at least report which pics are not
stiched and suggest to experiment with some settings
this is a good idea. I will implement this if I find some time, or you can do it and create a PR if you wish
from stitching.
lukasalexanderweber
commented on August 21, 2024
To summarize:
Mostly users of the stitching package will have a set of images which ALL should be included in the final panorama. By setting the confidence threshold to 1, by default images are sorted out if they don't have good enough matches. In this case the user should be informed.
from stitching.
lukasalexanderweber
commented on August 21, 2024
BTW do you need help for your images or have you figured out how to generate a complete panorama?
from stitching.
emard
commented on August 21, 2024
Of course I'd like some help with images
I can prepare some minimal set of images
that demonstrates the problem. Each image
is full HD 1920x1080 jpg, they overlap at least
30% left to right. Stitching about 40 of them
should produce about 40000x1080 image.
It's great if you can suggest solution I'd like to have
automatic script that would stitch such images
images are every time similar in color and content.
…On 10/19/22, Lukas Weber ***@***.***> wrote:
BTW do you need help for your images or have you figured out how to generate
a complete panorama?
--
Reply to this email directly or view it on GitHub:
#30 (comment)
You are receiving this because you authored the thread.
Message ID: ***@***.***>
from stitching.
lukasalexanderweber
commented on August 21, 2024
Yeah a Sample Set would be helpful
from stitching.
emard
commented on August 21, 2024
HI
Here is 25MB zip download link valid in next 10 days there are
pictures that I'd like to stitch automatically
with commandline I used and result when not all pictures are in.
I'd like to stich them all horizontally left-to-right, sorted filename order.
Camera optics is pretty good x/y grid even and rectangular, same distance
from camera to object, perpendicular to surface and no need for any fisheye
lens correction or similar.
images from this equipment will be a bit dark, slightly unsharp and with
similar texture. If I need to overlap more, I can make more pictures with
less camera moving step.
https://jumboiskon.tportal.hr/download/1e80f803-29f8-4ebf-ae12-c1e24241075a
…On 10/19/22, Lukas Weber ***@***.***> wrote:
Yeah a Sample Set would be helpful
--
Reply to this email directly or view it on GitHub:
#30 (comment)
You are receiving this because you authored the thread.
Message ID: ***@***.***>
from stitching.
lukasalexanderweber
commented on August 21, 2024
try
stitch Screen* --confidence_threshold 0.2 --matches_graph_dot_file matches.txt --detector sift
and add
--final_megapix 0.01
if you don't want to wait for so long if you're just trying out some settings.
--matches_graph_dot_file matches.txt
this will create a file with matching informations, especially the confidences. We can see that with the default settings the confidences are too low, thats why so many images are filtered out. We can lower the --confidence_threshold to include more images. Changing the Feature Detector to SIFT improves the matches significantly. Here is the resulting matching graph:
graph matches_graph{
"Screen20221018_132345.jpg" -- "Screen20221018_132408.jpg"[label="Nm=8, Ni=8, C=0.769231"];
"Screen20221018_132408.jpg" -- "Screen20221018_132445.jpg"[label="Nm=14, Ni=14, C=1.14754"];
"Screen20221018_132500.jpg" -- "Screen20221018_132515.jpg"[label="Nm=6, Ni=6, C=0.612245"];
"Screen20221018_132544.jpg" -- "Screen20221018_132610.jpg"[label="Nm=38, Ni=38, C=1.95876"];
"Screen20221018_132610.jpg" -- "Screen20221018_132628.jpg"[label="Nm=7, Ni=7, C=0.693069"];
"Screen20221018_132628.jpg" -- "Screen20221018_132646.jpg"[label="Nm=29, Ni=29, C=1.73653"];
"Screen20221018_132646.jpg" -- "Screen20221018_132702.jpg"[label="Nm=19, Ni=19, C=1.38686"];
"Screen20221018_132725.jpg" -- "Screen20221018_132745.jpg"[label="Nm=33, Ni=29, C=1.62011"];
"Screen20221018_132745.jpg" -- "Screen20221018_132806.jpg"[label="Nm=327, Ni=315, C=2.9689"];
"Screen20221018_132806.jpg" -- "Screen20221018_132825.jpg"[label="Nm=24, Ni=21, C=1.38158"];
"Screen20221018_132825.jpg" -- "Screen20221018_132839.jpg"[label="Nm=28, Ni=28, C=1.70732"];
"Screen20221018_132839.jpg" -- "Screen20221018_132911.jpg"[label="Nm=40, Ni=40, C=2"];
"Screen20221018_132925.jpg" -- "Screen20221018_132958.jpg"[label="Nm=24, Ni=24, C=1.57895"];
"Screen20221018_132958.jpg" -- "Screen20221018_133013.jpg"[label="Nm=41, Ni=39, C=1.92118"];
"Screen20221018_133013.jpg" -- "Screen20221018_133031.jpg"[label="Nm=22, Ni=21, C=1.43836"];
"Screen20221018_133031.jpg" -- "Screen20221018_133043.jpg"[label="Nm=39, Ni=39, C=1.9797"];
"Screen20221018_133043.jpg" -- "Screen20221018_133057.jpg"[label="Nm=8, Ni=8, C=0.769231"];
"Screen20221018_133218.jpg" -- "Screen20221018_133240.jpg"[label="Nm=41, Ni=41, C=2.0197"];
"Screen20221018_133240.jpg" -- "Screen20221018_133326.jpg"[label="Nm=28, Ni=24, C=1.46341"];
"Screen20221018_133326.jpg" -- "Screen20221018_133343.jpg"[label="Nm=40, Ni=34, C=1.7"];
"Screen20221018_133343.jpg" -- "Screen20221018_133352.jpg"[label="Nm=10, Ni=10, C=0.909091"];
"Screen20221018_133352.jpg" -- "Screen20221018_133417.jpg"[label="Nm=25, Ni=25, C=1.6129"];
"Screen20221018_133417.jpg" -- "Screen20221018_133438.jpg"[label="Nm=11, Ni=11, C=0.973451"];
"Screen20221018_133438.jpg" -- "Screen20221018_133453.jpg"[label="Nm=9, Ni=9, C=0.841121"];
"Screen20221018_133506.jpg" -- "Screen20221018_133519.jpg"[label="Nm=12, Ni=12, C=1.03448"];
"Screen20221018_133519.jpg" -- "Screen20221018_133536.jpg"[label="Nm=77, Ni=77, C=2.47588"];
"Screen20221018_133536.jpg" -- "Screen20221018_133558.jpg"[label="Nm=8, Ni=8, C=0.769231"];
"Screen20221018_133558.jpg" -- "Screen20221018_133613.jpg"[label="Nm=28, Ni=28, C=1.70732"];
"Screen20221018_133613.jpg" -- "Screen20221018_133623.jpg"[label="Nm=15, Ni=15, C=1.2"];
"Screen20221018_133640.jpg" -- "Screen20221018_133650.jpg"[label="Nm=61, Ni=49, C=1.86312"];
"Screen20221018_133132.jpg";
"Screen20221018_133631.jpg";
}
only two images are not processed, I havn't analized why. For a graphic representation visit graphviz
Here is a low resolution result of the stitching
from stitching.
emard
commented on August 21, 2024
HI
Thanks for the improvement! I have tried options you recommended
for other similar images done with 2x less magnification and it stitched
them all, compared to my trying when only half of images stitched.
I can experiment with more overlapping, of course if you find improvement
on the difficult set, I'd like to try it
Regards, Davor
…On 10/20/22, Lukas Weber ***@***.***> wrote:
try
`stitch Screen* --confidence_threshold 0.2 --matches_graph_dot_file
matches.txt --detector sift`
and add
`--final_megapix 0.01`
if you don't want to wait for so long if you're just trying out some
settings.
`--matches_graph_dot_file matches.txt`
this will create a file with matching informations, especially the
confidences. We can see that with the default settings the confidences are
too low, thats why so many images are filtered out. We can lower the
`--confidence_threshold` to include more images. Changing the Feature
Detector to SIFT improves the matches significantly. Here is the resulting
matching graph:
```
graph matches_graph{
"Screen20221018_132345.jpg" -- "Screen20221018_132408.jpg"[label="Nm=8,
Ni=8, C=0.769231"];
"Screen20221018_132408.jpg" -- "Screen20221018_132445.jpg"[label="Nm=14,
Ni=14, C=1.14754"];
"Screen20221018_132500.jpg" -- "Screen20221018_132515.jpg"[label="Nm=6,
Ni=6, C=0.612245"];
"Screen20221018_132544.jpg" -- "Screen20221018_132610.jpg"[label="Nm=38,
Ni=38, C=1.95876"];
"Screen20221018_132610.jpg" -- "Screen20221018_132628.jpg"[label="Nm=7,
Ni=7, C=0.693069"];
"Screen20221018_132628.jpg" -- "Screen20221018_132646.jpg"[label="Nm=29,
Ni=29, C=1.73653"];
"Screen20221018_132646.jpg" -- "Screen20221018_132702.jpg"[label="Nm=19,
Ni=19, C=1.38686"];
"Screen20221018_132725.jpg" -- "Screen20221018_132745.jpg"[label="Nm=33,
Ni=29, C=1.62011"];
"Screen20221018_132745.jpg" -- "Screen20221018_132806.jpg"[label="Nm=327,
Ni=315, C=2.9689"];
"Screen20221018_132806.jpg" -- "Screen20221018_132825.jpg"[label="Nm=24,
Ni=21, C=1.38158"];
"Screen20221018_132825.jpg" -- "Screen20221018_132839.jpg"[label="Nm=28,
Ni=28, C=1.70732"];
"Screen20221018_132839.jpg" -- "Screen20221018_132911.jpg"[label="Nm=40,
Ni=40, C=2"];
"Screen20221018_132925.jpg" -- "Screen20221018_132958.jpg"[label="Nm=24,
Ni=24, C=1.57895"];
"Screen20221018_132958.jpg" -- "Screen20221018_133013.jpg"[label="Nm=41,
Ni=39, C=1.92118"];
"Screen20221018_133013.jpg" -- "Screen20221018_133031.jpg"[label="Nm=22,
Ni=21, C=1.43836"];
"Screen20221018_133031.jpg" -- "Screen20221018_133043.jpg"[label="Nm=39,
Ni=39, C=1.9797"];
"Screen20221018_133043.jpg" -- "Screen20221018_133057.jpg"[label="Nm=8,
Ni=8, C=0.769231"];
"Screen20221018_133218.jpg" -- "Screen20221018_133240.jpg"[label="Nm=41,
Ni=41, C=2.0197"];
"Screen20221018_133240.jpg" -- "Screen20221018_133326.jpg"[label="Nm=28,
Ni=24, C=1.46341"];
"Screen20221018_133326.jpg" -- "Screen20221018_133343.jpg"[label="Nm=40,
Ni=34, C=1.7"];
"Screen20221018_133343.jpg" -- "Screen20221018_133352.jpg"[label="Nm=10,
Ni=10, C=0.909091"];
"Screen20221018_133352.jpg" -- "Screen20221018_133417.jpg"[label="Nm=25,
Ni=25, C=1.6129"];
"Screen20221018_133417.jpg" -- "Screen20221018_133438.jpg"[label="Nm=11,
Ni=11, C=0.973451"];
"Screen20221018_133438.jpg" -- "Screen20221018_133453.jpg"[label="Nm=9,
Ni=9, C=0.841121"];
"Screen20221018_133506.jpg" -- "Screen20221018_133519.jpg"[label="Nm=12,
Ni=12, C=1.03448"];
"Screen20221018_133519.jpg" -- "Screen20221018_133536.jpg"[label="Nm=77,
Ni=77, C=2.47588"];
"Screen20221018_133536.jpg" -- "Screen20221018_133558.jpg"[label="Nm=8,
Ni=8, C=0.769231"];
"Screen20221018_133558.jpg" -- "Screen20221018_133613.jpg"[label="Nm=28,
Ni=28, C=1.70732"];
"Screen20221018_133613.jpg" -- "Screen20221018_133623.jpg"[label="Nm=15,
Ni=15, C=1.2"];
"Screen20221018_133640.jpg" -- "Screen20221018_133650.jpg"[label="Nm=61,
Ni=49, C=1.86312"];
"Screen20221018_133132.jpg";
"Screen20221018_133631.jpg";
}
```
only two images are not processed, I havn't analized why. For a graphic
representation visit [graphviz](https://dreampuf.github.io/GraphvizOnline)
Here is a low resolution result of the stitching
--
Reply to this email directly or view it on GitHub:
#30 (comment)
You are receiving this because you authored the thread.
Message ID: ***@***.***>
from stitching.
lukasalexanderweber
commented on August 21, 2024
You could try the other two detectors brisk or akaze, but other than that I dont see that much potential for further optimization but more overlap. Cheers Lukas
from stitching.
Related Issues (20)
- How to stitch a floor image? HOT 1
- Manually enforce image matches and stitching sequence HOT 4
- Change homography plane to avoid miniaturization effect
- Stitch works not correctly HOT 1
- Can I improve speed calculation ?
- How can I fix "On entry to DLASCLS parameter number 4 had an illegal value" HOT 7
- Retrieving Transformation Matrix Used to Stitch Mosaic? HOT 1
- Image Stitching angle issue HOT 1
- Question about stitching image follows longitude
- Regarding calculation of Homography matrix HOT 1
- Camera parameters adjusting failed. HOT 4
- store used parameters when using verbose mode HOT 1
- Request a complete configuration file (requirements.txt) that works HOT 4
- "Segmentation fault" error HOT 7
- Part of the stitching is missing HOT 1
- verbose mode shows matching images without matches when using `range_width`
- How to get an equirectangular image?
- Saving Plane warper points for video stitching HOT 1
- Images don't seem to stitch, despite pretty clear matches HOT 1
- The images are stitched longitudinally error HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from stitching.