0

I have two images, one is original image and another one is canny edge image. I like to remove noise actually from the original image. So I tried canny edge, it looks ok, but still have some noises. How can I remove those noises further in canny edge image or the main reason is just want to remove noise and keep those numbers. I found one discussion here. They suggest to use Bilateral filter or the Guided filter or BM3D filter before applying Canny edge. I applied bilateral filter before applying Canny edge. Still have some noises. How can I improve?
enter image description here

enter image description here

Community
  • 1
  • 1
batuman
  • 7,066
  • 26
  • 107
  • 229
  • Have you tried a slightly larger Gaussian kernel? – LSerni Mar 09 '14 at 15:43
  • I tried this org.opencv.core.Size size = new Size(15,15); Imgproc.GaussianBlur(roi_gray, roi_gray, size, 0.5); But still the same. – batuman Mar 09 '14 at 15:47
  • Sorry, I mean the kernel size you pass to Canny, which has a Gaussian convolution built-in. Another thing to try would be to convert to gray-intensity and apply a different (higher) thresholding. You may have to experiment with thresholding to see how it works out. – LSerni Mar 09 '14 at 15:50
  • The original image has 0 and 255 only. – batuman Mar 10 '14 at 00:46
  • Yes, since it was B/W, and *initially* you would have two levels in the converted image too. But when filtering, you'll get more shades, and it will be easier to distinguish between speckled noise and continuous signal. Try converting in gray in Windows Paint and resize to 50% size to see a *very* rough version of what I mean. – LSerni Mar 10 '14 at 12:34

0 Answers0