Closed mvcisback closed 10 years ago
http://en.m.wikipedia.org/wiki/Histogram_equalization It's to adjust contrast, it might not be as useful for a series of frames, but worth a shot.
On Friday, October 24, 2014, Marcell Vazquez-Chanlatte < notifications@github.com> wrote:
Brightness fluctuations in the scene cause superfluous changes between frames.
Please discuss and post results of relevant experiments here.
To combat this we've discussed a few options:
- Histogram Equalizing Need more information as to what this is
- Simple threshold that gradient must pass before being considered a change
- Changing color spaces and then projecting away the brightness information
- HSV
- YUV
- Sunglasses (also to demonstrate that gifs and images can be embedded)
— Reply to this email directly or view it on GitHub https://github.com/mvcisback/CS598ps_project/issues/4.
Thanks, Best Regards, Ramin
A good paper on image preprocessing. http://cmp.felk.cvut.cz/~flachbor/teaching/cv_int_rob/bright_geom_transf.pdf
@ramili I think it may be worth a shot.
exactly! comparing across scenes is going to be a problem. There might be ways to apply hist eq across all frame with respect to a ref frame. If you assume the lighting doesn't change drastically during the speech, then applying them individually should be okay too. Something to play with. On matlab you can use "histeq" to do that. http://www.mathworks.com/help/images/ref/histeq.html
We're going to use a face recognition system as a black box that presumably handles this. Thus closing
Brightness fluctuations in the scene cause superfluous changes between frames.
Please discuss and post results of relevant experiments here.
To combat this we've discussed a few options: