Edge detection with Sobel and Scharr kernels

Fig. 1: Original photo
Fig. 2: Photo convolved with a separable 3x3 Sobel kernel,
pixel intensities reduced to 1/2, inverted
Fig. 3: Photo convolved with a separable 3x3 Scharr kernel,
pixel intensities reduced to 1/9, inverted

The Scharr kernel gives more pronounced edges here, so it needs a stronger pixel intensity reduction for a comparable result. If the image convolved with the Sobel kernel was also with pixel intensities reduced to 1/9, the edges would become much less visible.

Fig. 4: Photo convolved with a separable 3x3 Sobel kernel,
pixel intensities reduced to 1/9, inverted