Identifying overlapping lines in greyscale images

Hi,

I've got two example images of a carbon fibre polymer from optical lockin-thermography at two different excitation frequencies.

In the first, you can clearly see the direction of the carbon fibre bundles in the first layer.

The second contains the same first layer lines, but also additional lines from the second layer, at a 45° angle to the original layer's.

I'm trying to write a script to automatically recognise the new layer line direction via Hough algorithm.

The problem is that the new layer lines are apparently too diffuse for the edge detection (via canny algorithm) to pick up on; the diagonal lines are suppressed and interrupted by the vertical lines.

Simply subtracting one image from the other doesn't improve the result measurably, there are still few to no diagonal lines being recognised by the Hough algorithm.

I'd be very grateful for any ideas on how to either improve my edge detection image, or other solutions for how to recognise the 2nd layer orientation.

I've attached the two phase images from my OLT for you all to play around with.