如何在三维数据立方体中使用列表推导式进行值过滤？

Hey there! Let's figure out how to extend that list comprehension to handle your 3D image cube.

First, let's recap your working 2D code for a single image:

AM2 = [[x if x > 1e-5 else 0 for x in line] for line in AM[0]]

This iterates over each line in AM[0] (your first image), then each pixel value x in that line, replacing values below 1e-5 with 0.

For a 3D cube (where AM is a list of 2D images), you just need to add one more outer loop to iterate over each image in the cube. Here's the modified code:

AM_filtered = [[[x if x > 1e-5 else 0 for x in line] for line in image] for image in AM]

Let's break this down layer by layer:

• The outermost loop for image in AM goes through every 2D image in your 3D cube.
• The middle loop for line in image processes each row (line) of the current image.
• The innermost loop for x in line checks each pixel value, replacing it with 0 if it's ≤1e-5, otherwise keeping the original value.

Example to visualize the structure #

Suppose your 3D cube AM looks like this (simplified with small values):

AM = [
    [[0.0, 1.1e-5], [2.2e-5, 0.0]],  # Image 1
    [[3.3e-5, 0.0], [0.0, 4.4e-5]]   # Image 2
]

After applying the filtered list comprehension, AM_filtered would become:

[
    [[0, 1.1e-5], [2.2e-5, 0]],  # Image 1 (0s where values ≤1e-5)
    [[3.3e-5, 0], [0, 4.4e-5]]   # Image 2
]

Bonus: Faster alternative with NumPy (if applicable) #

If you're working with large image datasets, list comprehensions can be slow. Using NumPy would make this operation much more efficient and concise:

import numpy as np

AM_np = np.array(AM)
AM_filtered_np = np.where(AM_np > 1e-5, AM_np, 0)

This vectorized operation does the same filtering in one line, and it's way faster for big 3D arrays.

Hope that solves your problem! 😊

内容的提问来源于stack exchange，提问作者Chicrala