Ask Your Question
0

Decreased performance (fps) when increasing camera resolution?

asked 2013-05-04 01:40:01 -0600

MattiasR gravatar image

I am currently working on a facetracking/eyetracking system to measure som parameters from the eyes of that person that is in front of the camera. It uses both detecting and tracking so that I am able to increase the number of fps. I am now up to 21fps while measuring some parameters of the eye. (1.3 ghz processor,so i think that it is quite good, processor=~40-60%) But the resolution of my camera is to low (the camera needs to be to close to the person), but before I buy a new cam with much higher resolution (>=1280x...) i want to ask here if there is someone who have done this before?

Do you think that my system will suffer, with a decreased fps, when increasing the resolution of the camera? While tracking the person/eyes i am only analysing a box of 100x50px, so not the whole cam frame.

I know that this is a very blurry question, but I still want to ask it, and hopefully someone can giver their opinion and thoughts. The cam is excpensive (>500$)

Thanks!

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
1

answered 2013-05-04 04:55:58 -0600

A larger resolution means basically more data to process. So to answer your question, yes it will slow down your algorithm because basically you still need to read in a complete frame and then process a region of interest, which means more data in the first step of acquiring frames.

However, some smart camera's allow you to read in a predefined region of interest of your actual camera CCD/CMOS sensor, but this only helps in specific cases where you know the actual region of processing beforehand.

Also, if your region of interest stays equal, but the amount of pixels inside it rises due to larger resolution, the process time of that region of interest will also go up. Keep this in mind.

Getting real time processing algorithms is always a tradeoff between resolution, fps, processing power, ... it is something you will need to dig into and experience through trial and error.

edit flag offensive delete link more

Comments

Thank you for your answer! I didn't think about the increased amount of data even when having a ROI of the matrix. But thats true! I wonder. Do you know if the two steps: grab a frame, and grab the ROI part of the frame, if the original is of high res?. Or is that just minor work compared to the tracking/detecting part?

MattiasR gravatar imageMattiasR ( 2013-05-04 14:08:49 -0600 )edit

I guess it depends on what high resolution means for you. I had a project with 25000 x 25000 pixel images, and those took 2-3 seconds to load completely. However, since for detecting/tracking, you can reduce your search space, it depends... if you need to search the complete image on all possible scales, then it will be the demanding time factor.

StevenPuttemans gravatar imageStevenPuttemans ( 2013-05-05 01:23:04 -0600 )edit

Question Tools

Stats

Asked: 2013-05-04 01:40:01 -0600

Seen: 1,010 times

Last updated: May 04 '13