random sampling pixels by MNIST

I am a beginner of deep learning with Tensorflow

Recently,I did image classification by tutorial MNIST.

Since Ｉ use a neural network which selects the pixel value of one pixel for input, I think that even if some pixels are sampled by random mask, it can recognize to some extent.

So, I want to randomly extract 100 pixels by masking x_image like this, but I can not use random mask well.

How can I sample 100 pixels without duplication from the input?

#import
from tensorflow.examples.tutorials.mnist import input_data
import tensorflow as tf
import numpy as np
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)

#modeling


x = tf.placeholder　#I want to load images

mask = np.array(["True:100 False 684"]) #I want to randomly add 100 elements of True

　　

x = boolean_mask(x,mask) #I want to extract 100 pixels with a mask

W = tf.Variable(tf.zeros([100,10]))
b = tf.Variable(tf.zeros([10]))
y = tf.nn.softmax(tf.matmul(x, W) + b)
y_ = tf.placeholder(tf.float32, [None, 10])

cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y), reduction_indices=[1]))
train_step = tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)

init = tf.initialize_all_variables()
sess = tf.Session()

sess.run(init)

#training
for i in range(1000):
  batch_xs, batch_ys = mnist.train.next_batch(100)
  sess.run(train_step, feed_dict={x: batch_xs, y_: batch_ys})

 correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))
 accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
 print(sess.run(accuracy, feed_dict={x: mnist.test.images, y_: mnist.test.labels}))