how to prepare images dataset for autoencoder

I used this link to train 20000 2D images with size 256x256. first I have to prepare the dataset to feed it into autoencoder fuction so i followed the same steps in the link

the inputwidth and inputhight are 256, input size is 65536 and xTrain= zeros (65536,20000)

during the training i always get " out of memory" so i had to reduce the number of images from 20000 into 15000 , 10000 the same error . The only number which the code accepted was 500 images.