Just to confirm, I in this slide is meant to mean Image and not the identity matrix.
ShivanPatel2025
How effective is sobel edge detection in cases with subtle contrast differences between edges?
aayushg55
@ShivanPatel2025 If the intensity change across an edge is very gradual, the gradient computed by the Sobel kernels will be relatively small, meaning edges with subtle contrast differences may appear faint in the output gradient magnitude image. Another weakness is that Sobel Edge Detection can be poor in the presence of noise, as the gradients in noise could overpower those of weak edges.
More robust edge detection techniques may be better suited depending on the use case, such as the Canny edge detector. This applies Gaussian smoothing to reduce noise before Sobel filtering along with non-maximum suppression.
jasonTelanoff
It makes sense to apply a smoothing to limit the effect of noise, but wouldn't that remove the chance of picking up on the more subtle between edges? Is there an algorithm that can somehow use the smoothed edges and non-smoothed edges to show all edges, even the faint ones, but no noise?
Rohan1215
While reading about popular filters, I came across a filter for making images or videos appear hand-drawn called the cartoon filter. It utilizes Sobel edge detection to identify edges and enhance them further. Then, it blurs and simplifies colors and details to simulate the less realistic hand-drawn effect. Lastly, it overlays the original image slightly.
Just to confirm, I in this slide is meant to mean Image and not the identity matrix.
How effective is sobel edge detection in cases with subtle contrast differences between edges?
@ShivanPatel2025 If the intensity change across an edge is very gradual, the gradient computed by the Sobel kernels will be relatively small, meaning edges with subtle contrast differences may appear faint in the output gradient magnitude image. Another weakness is that Sobel Edge Detection can be poor in the presence of noise, as the gradients in noise could overpower those of weak edges.
More robust edge detection techniques may be better suited depending on the use case, such as the Canny edge detector. This applies Gaussian smoothing to reduce noise before Sobel filtering along with non-maximum suppression.
It makes sense to apply a smoothing to limit the effect of noise, but wouldn't that remove the chance of picking up on the more subtle between edges? Is there an algorithm that can somehow use the smoothed edges and non-smoothed edges to show all edges, even the faint ones, but no noise?
While reading about popular filters, I came across a filter for making images or videos appear hand-drawn called the cartoon filter. It utilizes Sobel edge detection to identify edges and enhance them further. Then, it blurs and simplifies colors and details to simulate the less realistic hand-drawn effect. Lastly, it overlays the original image slightly.