Welcome to OGeek Q&A Community for programmer and developer-Open, Learning and Share
Welcome To Ask or Share your Answers For Others

Categories

0 votes
264 views
in Technique[技术] by (71.8m points)

python - Adding different sized/shaped displaced NumPy matrices

In short: I have two matrices (or arrays):

import numpy

block_1 = numpy.matrix([[ 0, 0, 0, 0, 0],
                        [ 0, 0, 0, 0, 0],
                        [ 0, 0, 0, 0, 0],
                        [ 0, 0, 0, 0, 0]])

block_2 = numpy.matrix([[ 1, 1, 1],
                        [ 1, 1, 1],
                        [ 1, 1, 1],
                        [ 1, 1, 1]])

I have the displacement of block_2 in the block_1 element coordinate system.

pos = (1,1)

I want to be able to add them (quickly), to get:

[[0 0 0 0 0]
 [0 1 1 1 0]
 [0 1 1 1 0]
 [0 1 1 1 0]]

In long: I would like a fast way to add two different shape matrices together, where one of the matrices can be displaced. The resulting matrix must have the shape of the first matrix, and the overlapping elements between the two matrices are summed. If there is no overlap, just the first matrix is returned unmutated.

I have a function that works fine, but it's kind of ugly, and elementwise:

def add_blocks(block_1, block_2, pos):
    for i in xrange(0, block_2.shape[0]):
        for j in xrange(0, block_2.shape[1]):
            if (i + pos[1] >= 0) and (i + pos[1] < block_1.shape[0])
               and (j + pos[0] >= 0) and (j + pos[0] < block_1.shape[1]):
                block_1[pos[1] + i, pos[0] + j] += block_2[i,j]
    return block_1

Can broadcasting or slicing perhaps do this?

I feel like maybe I'm missing something obvious.

See Question&Answers more detail:os

与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
Welcome To Ask or Share your Answers For Others

1 Reply

0 votes
by (71.8m points)

An easy solution that looks like MATLAB solution is:

import numpy as np

block1 = np.zeros((5,4))
block2 = np.ones((3,2))

block1[1:4,2:4] += block2  # use array slicing

print(block1)

[[0. 0. 0. 0.]
 [0. 0. 1. 1.]
 [0. 0. 1. 1.]
 [0. 0. 1. 1.]
 [0. 0. 0. 0.]]

So package it as a reusable function:

import numpy as np

def addAtPos(mat1, mat2, xypos):
    """
    Add two matrices of different sizes in place, offset by xy coordinates
    Usage:
      - mat1: base matrix
      - mat2: add this matrix to mat1
      - xypos: tuple (x,y) containing coordinates
    """
    x, y = xypos
    ysize, xsize = mat2.shape
    xmax, ymax = (x + xsize), (y + ysize)
    mat1[y:ymax, x:xmax] += mat2
    return mat1

block1 = np.zeros((5,4))
block2 = np.ones((3,2))
pos = (2,1)
print(addAtPos(block1, block2, pos))

[[0. 0. 0. 0.]
 [0. 0. 1. 1.]
 [0. 0. 1. 1.]
 [0. 0. 1. 1.]
 [0. 0. 0. 0.]]

与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
OGeek|极客中国-欢迎来到极客的世界,一个免费开放的程序员编程交流平台!开放,进步,分享!让技术改变生活,让极客改变未来! Welcome to OGeek Q&A Community for programmer and developer-Open, Learning and Share
Click Here to Ask a Question

...