How to Efficiently Convert PIL Images to NumPy Arrays for Pixel Transformations?

Converting PIL Images to NumPy Arrays for Efficient Pixel Transformations

When working with image processing tasks, it is often desirable to convert a PIL (Python Imaging Library) image into a NumPy array for faster pixel-wise transformations. This conversion enables efficient manipulation of image data, allowing for more complex and time-optimized image operations.

To convert a PIL image to a NumPy array, one can use the following code snippet:

pic = Image.open("foo.jpg")
pix = numpy.array(pic.getdata()).reshape(pic.size[0], pic.size[1], 3)

This code reads the image data from the PIL image and reshapes it into a 3-dimensional NumPy array, where each dimension represents the image's height, width, and channel (e.g., RGB).

To convert the NumPy array back into a PIL image after performing the desired transformations, the following code can be used:

new_pic = Image.fromarray(modified_pix)

However, this method does not allow for direct modification of the original PIL image using the modified NumPy array.

Prior to PIL version 1.1.6, modifying the original PIL image required converting the NumPy array to a list of tuples:

data = list(tuple(pixel) for pixel in modified_pix)
pic.putdata(data)

However, this approach can be slow and inefficient.

In PIL version 1.1.6 and above, the preferred method for converting between PIL images and NumPy arrays is:

pix = numpy.array(pic)

This command produces a 3-dimensional NumPy array with rows, columns, and RGB channels as its dimensions.

After performing the pixel transformations, the updated array can be converted back to a PIL image using:

new_pic = Image.fromarray(modified_pix)

Alternatively, one can modify the original PIL image using the modified NumPy array with:

pic.putdata(modified_pix)

The above is the detailed content of How to Efficiently Convert PIL Images to NumPy Arrays for Pixel Transformations?. For more information, please follow other related articles on the PHP Chinese website!