What's a good way to get into image programming?
Learning how jpg's and png's work and ways to create algorithms based on image contents. Stuff like that.
Just be yourself bro.
>>1480260
Is that cryptic for "just do projects you like"?
>>1480305
What?
>>1480255
If you want to know how the file formats specifically work, just read the specs. Of course, for most stuff you would do with images, how the pixels are compressed and stored isn't particularly important other than understanding what lossy/lossless is. Of course it's a good idea to learn what color spaces and transfer functions/gamma are so you have some idea what the pixels actually represent, but even that's an afterthought most of the time.
Probably the most important thing would be to get some library for working with images and start doing projects that interest you.
This is mostly about video (and audio), but some stuff is relevant even to static images. You might find the video stuff interesting too, though.
https://www.xiph.org/video/vid1.shtml
I don't understand why some people still come here to make serious threads.
>>1480374
Because there are still people who come here to post serious replies?
/qa/ has always had tons of spammers and shitposters - the only difference is that the current crop doesn't try to pretend that they're spammers and shitposters, but that's offset by their much lower numbers compared to almost any point in the last year or so. This is nothing new.
pretend that they're not*
>>1480359
Thanks for the serious response
>>1480255
OpenCV. It's a free computer vision library written in C++ and Python. You can use feature detection to train image sets. Real-time detection is also possible if you use a video stream. It's pretty robust as template matching and SIFT algorithms are built in.
>>1480255
This >>1480571 but it depends on your expertise
If you've done nothing at all like that, I'd suggest you begin with something simpler, so that you have to implement algorithms yourself.
At the very least, start with Python and Numpy + Pillow. They're excellent together, and Numpy is quite fast for simple tasks (IOW, things that don't require a GPU). Don't underestimate it: if you feel that an operation on an array (say) should be a single method call, it probably is. Plus it's literally APL + Python so there's that.
If you're happy with that, good. If what you wanted to do was stuff like feature detection, things like NNs, feature detection, recognizing images and the like, then OpenCV and specialized NN libraries are your best friends, because they're really, really fast, and they make most things much easier.
>>1480639
This
>>1480639
>Don't underestimate it: if you feel that an operation on an array (say) should be a single method call, it probably is.
To elaborate on this: Telling the numpy library to do an operation on the whole array at once is much faster than iterating over the whole array in Python. You should write code that takes advantage of this speedup.
keeping up