This will highly depend on the nature of your images and the application you're looking for. However, as soon as you have segmented your image using Otsu's method or any other thresholding method, I think there is no need for further edge detection algorithms. Detecting regions (or edges) in a binary image can be done easily, for instance using the MATLAB function "bwboundaries".

It strongly depedends on the context.

The Canny Edge Detector usually finds many more edges

than needed. Intelligent removal must take place.

One idea is that you make use of patterns to the

right and to the left of an edge in order to follow

it. Another issue is that edges might be interrupted

by dark or bright spots. So you have to bridge the

gap by exploiting the mathematical description

of the curvature. In the easiest case you have

a straight bone, then you're able to bridge an

interruption with polynomial extrapolation.

Regards,

Joachim

As pointed out above- no need for edge detection as it is apparent that you've already binarised the image with Otsu's threshold. Check out the below link for examples: http://www.mathworks.in/help/images/ref/bwboundaries.html

Canny is not suitable for edge detection in all cases and again it is not necessary to use canny after otsu is performed which is pointed out by the previous researchers.

Canny depends on the image u have.If it is retinal then it is fine.But after thresholding edge detection may not be necessary.U can go for some morphological operators like dilation,erosion etc,to fill the gaps.

You can try subpixelEdges in Matlab to obtain a more accurate subpixel edge detection in your image.

http://es.mathworks.com/matlabcentral/fileexchange/48908-accurate-subpixel-edge-location