Sobel on both X and Y

[Turtyo]

Description

This pull request implements a sobel filter using the magnitude of the X and Y components of the sobel filter to build the image. It calculates $\sqrt{X^2 + Y^2}$ for each pixel of the image, with X the horizontal sobel component, and Y the vertical sobel component.

Tests

It was discussed in the related issue ( closes #192 ) that tests were written in tests.rs, but calculating the expected result would just come to applying the algorithm by hand. As I noticed that none of the other convolution functions had a test, should I still write a test for it ?

Result

I modified src/bin/bin.rs quickly to test the result:

Original image:

Global sobel:

For comparison, here are both the horizontal and vertical sobel:

Horizontal	Vertical

Time

As a oneshot (so not very reliable but that gives an idea):

• Global: 66.7 ms
• Horizontal: 29.2ms
• Vertical: 29.7ms

The results are quite logical, since the global needs to calculate both the X and Y components (which should take around 60ms on my computer); a bit surprising that it only takes 6ms then to calculate the magnitude for each pixel, but since it was a single time test, maybe not the best to calculate things on.

PS: Those times are with --release flag

[Turtyo]

Not sure if you want to be pinged for PR to see them from the notification tab @silvia-odwyer but just in case I will ping

[silvia-odwyer]

@Turtyo Thanks very much for your excellent contribution, it's much appreciated! 🎉 I'm delighted that the library now has a global sobel function ✅