Canny edge detection is accomplished using the edge function. Matlab will get "automatic/adaptive" thresholds if you don't specify thresholds as an argument. An easy way to implement this is to simply call Edge twice; first to get an initial capture of the thresholds, then to adjust thresholds as needed based on the automatic thresholds.
[~,threshOut = edge(I,'Canny');
threshold = threshOut*1.0;
BW1 = edge(I,'Canny',threshold);
Just adjust the multiplier to adjust thresholds. Matlab's adaptive technique is fairly straightforward, and I believe it is a variant of the Otsu method; they simply throw out 70% of the detected edge pixels for the high threshold, and the low threshold is some constant factor multiplied by the high threshold.
There are a number of other self-adaptive ways to choose Canny thresholds, a few are detailed here: http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=5476095
