Please check my workflow: brightness verse time plot

asked 2014-09-06 10:05:52 -0600

loveandkindness gravatar image

updated 2014-09-06 10:10:50 -0600

Imagine that we have a image sequence of some Christmas tree lights stretched out on the ground. The lights are blinking away. We're not interested in colors, only brightness and periods.

Assuming that the lights don't move throughout the video, this is how I would tackle the problem:

  1. I would begin by using region of interest manager in ImageJ to quickly bound each light with a rectangle, then export the xy coordinates into a file.
  2. Next, load the xy coordinates into a Python script. Each light is to be cropped out of the large image and stored in a list of smaller images.
  3. On each image, I would then perform a gray scale conversion, Gaussian Blur, and Otsu's Binary Thresholding.
  4. Lastly, I would sum over the binary images to yield an integer, which should give a relative brightness. Store and loop for each image in the sequence, and for each light in each image.

Thank you very much for any advise or critique!

edit: Also, imagine that someone will be randomly shining a flashlight at different lights in the video. Any ideas on reducing the damage from this, if possible, would be wonderful!

edit retag flag offensive close merge delete

Comments

I just found an ImageJ plugin (http://rsb.info.nih.gov/ij/plugins/time-series.html) that does this entire process, and outputs the results into a text file. I'm still very curious how others would handle this job, since the first method I come across can't always be the best.

loveandkindness gravatar imageloveandkindness ( 2014-09-06 12:18:53 -0600 )edit