How can I input two images into neural network?

In my project, the input to the Neural net. should be two images, one 256*256 the other 64*64.

p={ image1 image2} net=train(net, p, output}

note that the two images should be the same size. so what is the best way to inter these matrices?

is it to unroll these matrices into one big column ( then the number of neurons will be 69632)?! or, segment every picture into a number of blocks x*x pixels, then the unroll this block , then the neural net will treat every block alone?!

What is the data structure will be?

• " Concurrent Inputsin a Static Network ”or
• " Sequential Inputs in a Dynamic Network ”or
• " * Concurrent Inputs in a Dynamic Network "
• "Batch" or
• "Sequential" ?

thank you in advance

Answer 1

[Greg Heath]

Features should be extracted from each image. The features should be assembled into a vector. The type of net should be static concurrent.

Thank you for formally accepting my answer

Greg

[Alaa]

thank you for your reply my work is in Steganography,

I wrote this code input (Rs_1) represent the first block in secret image the others represent blocks in secret image =16 bock

output should be the same as cover image blocks = 16 block

p= [Rs_1 Rc_1 Rc_2 Rc_3 Rc_4 Rc_5 Rc_6 Rc_7 Rc_8 Rc_9 Rc_10 Rc_11 Rc_12 Rc_13 Rc_14 Rc_15 Rc_1 ];

t= [Rc_1 Rc_2 Rc_3 Rc_4 Rc_5 Rc_6 Rc_7 Rc_8 Rc_9 Rc_10 Rc_11 Rc_12 Rc_13 Rc_14 Rc_15 Rc_16 ];

% Train ANN net = feedforwardnet(17,'trainrp'); net.trainParam.epochs = 100; net.trainParam.show = 10; net.trainParam.goal = 0.001; net.performFcn='msereg'; net.performParam.ratio=0.5; net.divideParam.valRatio=0; net.trainParam.min_grad=0;

% Train ANN net=train(net,p, t);

a=net(p)

when I run this code, error declared that number of input and output should be the same

any help please