for any single image,we can detect the vein,but for lots of people,many veins in the arm cannot be picked up.I don't know how to improve the accuracy rate.we work hard to make the light stable. when the camera moved, the light changed.Thus the imaging process differs。
In order to help i need to know what is the process that you are doing now in order to detect the vein.
If you think the different lighting was the problem than try to use the gradient image and not the image itself for the detection process.
Dear Wenzhi Wang,
you could apply techniques like (local) automatic histogram equilization to normalize brightness and contrast in the image.
Also, make sure that all automatic settings of your camera (white balance, shutter, gain, etc.) are turned off, i.e., set to manual.
Cheers,
Jochen
Maybe is feasible for you to use infrared images? They're invariant to illumination variations, and you could detect the arm veins using thermal gradients caused by blood perfusion.
thank you all for answering my question. I don't know how to communicate with each single expertise.
The process is like this:downsample the original image, foreground segmention, vein enchancement(gabor filter), binaryzation，find the proper vein.
As Jochen Kerdels said, the camera's shutter and gain should be set invariant.Or the noise become large when the light changes.
Later on, I will try the gradients of the image.However, I am not that clear of the local automatic histogram? Is there any material about it ?
Dear Camilo Rodriguez, is the thermal gradients similar with the regular gradients of the image??
Dear Wenzhi Wang,
I found a nice article on histogram equilization on the OpenCV-Website:
http://docs.opencv.org/2.4/doc/tutorials/imgproc/histograms/histogram_equalization/histogram_equalization.html
Cheers,
Jochen
Yes, the thermal gradients are regular gradients of the image, but I use that term because they're produced by skin surface temperature variations, which are not exploitable in visible images.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing