Each byte should be 8 bits (0-255). If a 16-bit integer is stored/transferred using two bytes, then to "unpack" the 2 x 8 bits into 16 bits, you would do something like:
value = (byte_low | (byte_high << 8))
I don't know if the EV3 software allows you to use bitwise math. It's probably easier to do something like:
value = (byte_low + (byte_high * 256))
If value needs to end up being a 16-bit signed value (as opposed to unsigned), then in C you could type cast it to a int16_t, but in EV3 software you would probably need to do something like:
if value > 32767
value = value - 65536
Which byte is high and which is low is dependent on the device you're trying to read from. The device datasheet should tell you how the registers are configured.