If these shades of gray information are binarized, the main structure information of original image is completely lost. In order to deal with this problem, we propose to use the noise attack
In this part, why are you assuming gray information are binarized ? After output pass the sigmoid, it conatain the value between 0~1, not 0 and 1 (binary).
The information encoded with 0 and 1 is easier to resist the
noise attack.
In this part, why the information (output of ComCNN) encoded with 0 and 1 is easier to resit the noise attack ?
If anyone can give further explanation about this, it will really helpful for me.
In your paper, I have the following questions.
If these shades of gray information are binarized, the main structure information of original image is completely lost. In order to deal with this problem, we propose to use the noise attack
In this part, why are you assuming gray information are binarized ? After output pass the sigmoid, it conatain the value between 0~1, not 0 and 1 (binary).
The information encoded with 0 and 1 is easier to resist the noise attack.
In this part, why the information (output of ComCNN) encoded with 0 and 1 is easier to resit the noise attack ?
If anyone can give further explanation about this, it will really helpful for me.
Thanks.