questionable cuda kernel performance

I’m new to cuda and I’ve only learnt some basics about it
I recently wrote a code for a simple blur filter and another code for fuzzy based segmentation(which include some per-element multiply, add, divide and some summation which I wrote it based on samples for dot-product and a somewhat a blur-filter) but in first case my code was a bit slower!!! than CPU code and in the latter I only got 2x speed up.
I’ve heard cuda can speed up much higher but I don’t see any simple way to speed things up.
my CPU is Intel core i7 4700 HQ 2.4GHz and my GPU is GeForce GT 750m (I have N46JV Asus laptop)

is my results reasonable or I’m doing something wrong