bring CNN's Convolution process to After_imaging method 

We should first make 3 filters(which is same as CNN's convolutional process), and process image with these filters. 
And do the same thing as I did along, But this time with 3 Mem-crossbar architecture.
outputs from three crossbar would be added. 
This is for enhancing its ability in more non-linear situations. 