Hey /g/
I have a setup where I record cells going through a CMOS camera that I record with.
I basically want to plot how the intensity of the cell v time. I have frames where the cell is moving labelled 1-4, and when I draw a line segment and plot it against time, I get the spikes where I want a nice curve.
I get that there is a time delay between each frame, and the frame represents distance at a time rather than time itself, but I'm not sure how to code to get the curve I want.
Any suggestions would be greatly appreciated. Thanks!
I had an article on how to generate perlin noise that also had a nice tutorial on how to smooth out line graphs. Let me find it. Can you read C?
>>54462018
Yeah, I can read it. I am looking more for a concept on how to connect the streaks together to get a coherent curve.
I can code it in Matlab but I can't grasp the concept in my head for some reason...
>>54461715
I think you've stumbled upon the differences between continuous and discrete time.
Your problem is that you're sampling, which produces a discrete signal, while you're looking to get a continuous signal out the end. The best way to do this would be to find the highest frequency you want to measure, which will relate to the speed of your object and desired precision. Say you want to be able to pick up a 1 pixel change in an object moving 100 pixels/sec, you would have a frequency of 100Hz, or 100 samples per second, thugh this will yield a signal that is still in the discrete time domain. Once you figure that frequency out, double it to reach the Nyquist frequency. In the same example as above, at a 200Hz sampling rate you would be able to perfectly reproduce the original signal in continuous time from your discrete samples.
Look in to articles about moving from discrete to continuous time. I would just tell you how to do it, but my background is in the mathematics and analog signal processing, not digital or anything CS.
Best of luck
>>54461715
i don't but i would suggest you to read on how video encoder work. maybe you will find something there as
>>54462106
said
>>54461715
brah just use fft and get rid of the high f components (ie. low pass filter that shit). a bandpass filter could work as well but I'd need more details about how you're using the signal to be able to say anything about that.
>>54462667
I get the Nyquist theorem but I am not undersampling.
I am sampling the same cell in four different frames and I need a way to stitch it back together. Or I need to treat the cell as one object rather than an object in four separate frames... Basically, I need a way to "stitch" back the four frames.
The problem is, when I start having multiple cells at once, the stitching does not work.
>>54463799
FFTs of the high f component won't really help here since the the space for each bead is separated in time due to the frame capturing separate points in the position of the cell...
Basically I want the four separate streaks to come together like this but the problem comes when I have multiple beads lined up...
Here's the simplest case
>>54464569
Also here's the next step after...
