I am looking for an edge detection algorithm that works on images corrupted by severe Gaussian noise (sigma=20% or above). I came across Canny edge detector, bilateral filtering, and anisotropic diffusion in the literature, but I am not certain if these algorithms work on severely noisy edge detection. Can anyone please advise any algorithm that suits this situation?
Hi,
I've first applied our non-local-means filter [1] for imagej and get the denoised image rg_result.png (attached).
Next I've applied the common edge detection as implement in imagej [2].The final result is rg_result_edge.png (attached).
[1] http://imagej.net/Non_Local_Means_Denoise
[2] https://imagej.nih.gov/ij/docs/guide/146-29.html#toc-Subsection-29.3
Hi John and thanks so much for your response.
Just let me add that my final goal is not edge detection rather it is segmentation. Am keen to use edge enhancement in wavelet domain alongside other techniques to boost up the accuracy. I have attached two sample images.
Best,
Saeed
Morphological filters are powerful tools and can accomplish various
desired tasks: noise removal, edge detection, image smoothing, bilevel
image shape smoothing.
Based on my experience, Canny edge detector could do the job for sigma=20% or above. You might need to adjust the parameters for the canny operator.
Hi,
I've first applied our non-local-means filter [1] for imagej and get the denoised image rg_result.png (attached).
Next I've applied the common edge detection as implement in imagej [2].The final result is rg_result_edge.png (attached).
[1] http://imagej.net/Non_Local_Means_Denoise
[2] https://imagej.nih.gov/ij/docs/guide/146-29.html#toc-Subsection-29.3
Hi Saeed,
Sorry for my delay. I have used MIPAR (http://MIPAR.us) to make 2 recipes. One to find the noisy shapes in the first image, the other to find the object, foreground, and background in the second image. I have attached screenshots of the results, as well as the two recipe files for you to load into MIPAR to do the jobs automatically. Please contact me any time with any questions. The screenshots of the Image Processor in both cases reveal the custom, graphically-built algorithms (Recipes) to give you an idea of the steps used.
Cheers,
John
Please look into our article " Edge detection using fractional derivatives and Information sets" for a new approach where we also address the issue of noisy conditions.
Article Edge detection using Fractional derivatives and Information sets
there are several method one of them was edge detection, Discrete wavelet Transformation
Under such conditions, it is necesary to use a pre-processing stage. Maybe a median filter.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Algorithms