我想使用Python将32 * 32 * 3 rgb图像转换为32 * 32 * 1维度的灰度

Question

The following is my function to convert RGB to gray scale image.

My input image is 32*32*3 where as the output dimension looks like 32*32 , But i am looking for 32*32*1 . Do I need to resize or re scale this image.

Any thoughts?

def rgb2gray(rgb):
    return np.dot(rgb[...,:3], [0.299, 0.587, 0.114])

Answer 1

If you want the conversion to happen in the Tensorflow graph itself you can use this function: https://www.tensorflow.org/api_docs/python/tf/image/rgb_to_grayscale

tf.image.rgb_to_grayscale(input_images)

Also, looks like you are answering your own question. What is wrong with

def rgb2gray(rgb):
  return np.dot(rgb[...,:3], [0.299, 0.587, 0.114])

Good luck!

Answer 2

Many beginners have difficulty to understand this. Simply I'll explain you. If you have 20elements. You can convert them into ndarray of dimension [20] or [10,2] or [5,2,2]. So in your case, there are 32*32 = 1024 pixel data as your output. You can reshape them back using Numpy's reshape function. It's pretty simple.

For example,

print output.shape
// This prints as (32,32) in your case
output = output.reshape(32,32,1)

This solves your problem.