ME565 Lecture 18: FFT and Image Compression
ฝัง
- เผยแพร่เมื่อ 9 ก.ค. 2024
- ME565 Lecture 18
Engineering Mathematics at the University of Washington
FFT and Image Compression
Notes: faculty.washington.edu/sbrunto...
Matlab code: faculty.washington.edu/sbrunto...
Course Website: faculty.washington.edu/sbrunto...
faculty.washington.edu/sbrunton/ - วิทยาศาสตร์และเทคโนโลยี
The last part indeed blew my mind!! It was a beautiful explanation of how image compression is done. Thank you for making quality education freely available, Prof. Brunton!
Wow ! Thanks for the great lecture Steve. The last part totally blew my mind !!
Great lecture, thank you!
Nice explanation and very useful
This was brilliant. Thank you so much
38:50 - 43:08
You just got a new subscriber my friend.
Thanks!
Really great lecture, enjoyed the last compression concept!!! I have just one question if I want to do it on an RGB image instead of gray, how should I define threshold for loop when its all 3 rgb components.
Wow, excellent lecture!
Thank you!
Glad you liked it!
Thanx great lecture
What I do not understand is why the final image will weight less in sense of bytes if you will save it as a picture and look how much it weight in windows. I tried it and didt work' final image weighs same, and it makes sense as
you still have same size matrix, still each pixel has its own value. The FFT trashold you have done just smotthens the differences between pixels values, but still there is same amount of pixels as before and each has own value...
If you do not compress it, it will weight the same. But if you do compress it, then the computer would actually store just the high magnitude FFT values (along with where they occur) rather than the pixel values. This way it can store much less data and would transform back to the pixel data once you double click the image to open it.
@@anantchopra1663 I am sorry, guess I miss something. I thought that the very process described in this movie is the compression itself. Guy says he tresholds to zero a lot of information. So ghis is the compressing, isnt it?
@@igorg4129 Yes! When you set values less than a threshold to zero, that is the compression part, and _then_ you will save space.
@@anantchopra1663
So I tried, did same thing, and it doesnt. File weights almost same
Why do we use FFT?
Good question. I just did a whole series on this: th-cam.com/video/jNC0jxb0OxE/w-d-xo.html
There are a lot of reasons. One, sines and cosines have very intuitive interpretations for humans in terms of geometry and circles. They are also the building blocks for the solutions of linear differential equations, which come up all over the place. So representing signals as sums of sines and cosines is useful.
The Fourier basis also is useful for solving the heat equation and many other problems in scientific computing. The FFT allows us to solve these equations very quickly, so that we can investigate really complex scientific and engineering problems in simulations.
Also, the FFT is very useful for compressing data. So every time you watch a video on youtube or send a picture over your phone, you are using some algorithm related to the FFT.
@@Eigensteve it was helpful