8 ms·

Modern Image Processing Algorithms Implementation in C

199 points by symisc_devel 3y ago

Solvency 3y ago
Is there anything substantially different here from the way any modern post-processing shader works right now? For example: https://threejs.org/examples/webgl_postprocessing_rgb_halftone.html https://threejs.org/examples/webgl_postprocessing_rgb_halfto...
- dima55 3y ago
  Don't know about the algorithms, but this is a C library: for computers. threejs is a javascript library: for browsers
cwoolfe 3y ago
Visit the url below for much of the same content but without ads. https://sod.pixlab.io/ https://sod.pixlab.io/
Iwan-Zotow 3y ago
better than OpenCV ?
- realPtolemy 3y ago
  Curious to know as well
- slicktux 3y ago
  Better in what sense?
  - Iwan-Zotow 3y ago
    I mean OpenCV is C library, lots of well tested algorithms, Python binding, lots of users, docs, examples, you name it To be considered as potential use, it should either do something which OpenCV is not doing, or do things (a lot) better. I mean, I used OpenCV, shall I move to that lib? OpenCV is pretty much standard C imaging library, anyone working with images knowns about OpenCV
    - kilpikaarna 3y ago
      OpenCV is C++ (though there are 3rd party C bindings)
- cozzyd 3y ago
  easier to integrate into C-based stuff, probably...though unfortunately doesn't seem to make it easy to work without dynamic allocation, which makes it not as useful as it could be...
- spookie 3y ago
  The focus seems to make a smaller (not better) OpenCV
HumblyTossed 3y ago
Anything like this but for photo imaging manipulation?
realPtolemy 3y ago
Thanks for sharing!
ladberg 3y ago
Since it seems like the poster works for the company: the ads on this article are all semi-nsfw for me, more-so than the average GAd I see (is that something you can tune?). I reported them to Google but that might be a black hole. Additionally, why in the world does a technical blog for a for-profit company need ads?
- nico 3y ago
  Same here I got three images, two of pretty college women, and a third one that I can’t tell exactly what it is, but it’s skin color and my first impression was very NSFW
- symisc_devel 3y ago
  Hello, It's a blog post by one of our engineers. Authors are free to monetize their content without inference from the company.
  - chris37879 3y ago
    That sounds like a smart way to get your engineers to take the time for writing! I do assume you all do at least a basic suitability and security review, though?
    - symisc_devel 3y ago
      Of course. Only Adsense is authorized. After a short investigation, it appears that the main reason these NSFW-limit ads are shown is because the article includes direct link to the PixLab NSFW API Endpoint (https://pixlab.io/cmd?id=nsfw https://pixlab.io/cmd?id=nsfw) which is basically a bridge to our ML model that let you detect whether an image or video frame contains adult, bloody or gore content.
      - chris37879 3y ago
        Sorry if that came across like I was questioning that part of your process, I was more curious in the actual process, but it is neat that you guys already managed to figure that out!
elcritch 3y ago
Note this is GPL which will make it a harder sell vs OpenCV in many cases. I'd guess they're looking to use the non-GPL for $$ model. Nevertheless, it looks interesting. It looks really nice for embedded where you don't always have the GPU for compute.
remipch 3y ago
It looks promising. I like the way each algorithm is explained with minimal example and demo images. I'm looking for such a lightweight vision lib to embed simple image manipulation programs on tiny ESP32-CAM boards. OpenCV seems too heavy to integrate on such small devices. So far, I've been able to develop simple image processing programs with CImg[1] (simple filtering, image cropping, adding text). I'll try to spend some time exploring this SOD lib, thank you for sharing. [1] https://cimg.eu/ https://cimg.eu/
vitorsr 3y ago
How does it compare to dlib [1], ccv [2] etc.? [1] https://github.com/davisking/dlib https://github.com/davisking/dlib [2] https://github.com/liuliu/ccv https://github.com/liuliu/ccv
- liuliu 3y ago
  I like dlib better if you want a swiss army knife of algos. ccv is moved on to be all-in on neural-network based computer vision and all development effort happened in that area. (Similar can be said for dlib but at least the author still implement random black-box optimizers from time-to-time). SOD seems try to reimplement some of OpenCV but TBH, these are not "modern" at all. These are 15 year-old algos while works, the useful domain is shrinking and people who interested in these also might be looking for hardware-accelerated options (AVX-512 / CUDA etc). Some very specific classic heavy CV applications (such as SLAM) are quickly moving to their own dedicated libs (such as Cartographer from Google or ORB-SLAM).

199 points by symisc_devel 3y ago

Is there anything substantially different here from the way any modern post-processing shader works right now? For example: https://threejs.org/examples/webgl_postprocessing_rgb_halftone.html https://threejs.org/examples/webgl_postprocessing_rgb_halfto...

dima55 3y ago
Don't know about the algorithms, but this is a C library: for computers. threejs is a javascript library: for browsers

cwoolfe 3y ago

Visit the url below for much of the same content but without ads. https://sod.pixlab.io/ https://sod.pixlab.io/

Iwan-Zotow 3y ago

better than OpenCV ?

realPtolemy 3y ago
Curious to know as well
slicktux 3y ago
Better in what sense?
- Iwan-Zotow 3y ago
  I mean OpenCV is C library, lots of well tested algorithms, Python binding, lots of users, docs, examples, you name it To be considered as potential use, it should either do something which OpenCV is not doing, or do things (a lot) better. I mean, I used OpenCV, shall I move to that lib? OpenCV is pretty much standard C imaging library, anyone working with images knowns about OpenCV
  - kilpikaarna 3y ago
    OpenCV is C++ (though there are 3rd party C bindings)
cozzyd 3y ago
easier to integrate into C-based stuff, probably...though unfortunately doesn't seem to make it easy to work without dynamic allocation, which makes it not as useful as it could be...
spookie 3y ago
The focus seems to make a smaller (not better) OpenCV

HumblyTossed 3y ago

Anything like this but for photo imaging manipulation?

realPtolemy 3y ago

Thanks for sharing!

ladberg 3y ago

Since it seems like the poster works for the company: the ads on this article are all semi-nsfw for me, more-so than the average GAd I see (is that something you can tune?). I reported them to Google but that might be a black hole. Additionally, why in the world does a technical blog for a for-profit company need ads?

nico 3y ago
Same here I got three images, two of pretty college women, and a third one that I can’t tell exactly what it is, but it’s skin color and my first impression was very NSFW
symisc_devel 3y ago
Hello, It's a blog post by one of our engineers. Authors are free to monetize their content without inference from the company.
- chris37879 3y ago
  That sounds like a smart way to get your engineers to take the time for writing! I do assume you all do at least a basic suitability and security review, though?
  - symisc_devel 3y ago
    Of course. Only Adsense is authorized. After a short investigation, it appears that the main reason these NSFW-limit ads are shown is because the article includes direct link to the PixLab NSFW API Endpoint (https://pixlab.io/cmd?id=nsfw https://pixlab.io/cmd?id=nsfw) which is basically a bridge to our ML model that let you detect whether an image or video frame contains adult, bloody or gore content.
    - chris37879 3y ago
      Sorry if that came across like I was questioning that part of your process, I was more curious in the actual process, but it is neat that you guys already managed to figure that out!

elcritch 3y ago

Note this is GPL which will make it a harder sell vs OpenCV in many cases. I'd guess they're looking to use the non-GPL for $$ model. Nevertheless, it looks interesting. It looks really nice for embedded where you don't always have the GPU for compute.

remipch 3y ago

It looks promising. I like the way each algorithm is explained with minimal example and demo images. I'm looking for such a lightweight vision lib to embed simple image manipulation programs on tiny ESP32-CAM boards. OpenCV seems too heavy to integrate on such small devices. So far, I've been able to develop simple image processing programs with CImg[1] (simple filtering, image cropping, adding text). I'll try to spend some time exploring this SOD lib, thank you for sharing. [1] https://cimg.eu/ https://cimg.eu/

vitorsr 3y ago

How does it compare to dlib [1], ccv [2] etc.? [1] https://github.com/davisking/dlib https://github.com/davisking/dlib [2] https://github.com/liuliu/ccv https://github.com/liuliu/ccv

liuliu 3y ago
I like dlib better if you want a swiss army knife of algos. ccv is moved on to be all-in on neural-network based computer vision and all development effort happened in that area. (Similar can be said for dlib but at least the author still implement random black-box optimizers from time-to-time). SOD seems try to reimplement some of OpenCV but TBH, these are not "modern" at all. These are 15 year-old algos while works, the useful domain is shrinking and people who interested in these also might be looking for hardware-accelerated options (AVX-512 / CUDA etc). Some very specific classic heavy CV applications (such as SLAM) are quickly moving to their own dedicated libs (such as Cartographer from Google or ORB-SLAM).