Best possible technique to extract dimensions of a box from an image.

Hi, I am working on a project in which I have extract out dimensions(length, width and height) of a box from its picture. For that I applied edge detection filter (sobel filter using edge command), But it gives too much disconnected lines at the edges. I apply imdilate to remove those discontinuities but with that I am afraid it wont give the correct dimensions. So I needed guidance what would be a good starting point i.e which process would result in better extraction of dimensions. Thanks you.

[David Young]

The canny edge detector sometimes gives better results than sobel. (It's an option to edge.) But you may still have the same problems. Maybe you could fit straight lines to the edges, using the hough transform? It's hard to say more unless you post the image.

[Image Analyst]

You forgot to attach your image. Everyone, for some reason, thinks that edge detection is always the first step in image segmentation, when often/mostly it's not. Perhaps simple thresholding will work fine, but I don't know because I haven't seen your image.

[Maheen]

Here is the picture

Answer 1

[Walter Roberson]

The best possible technique starts with creating customized universes, and evolves from there. Are you certain that you have the resources to implement the best possible technique?

[Maheen]

I certainly don't intend to introduce one, I am trying apply any of the available techniques/methods. So I want to have an idea what would be the best starting point.