如何在python中将另一个图像居中成零矩阵？

Question

我是python图像处理的新手，我正在尝试为基于图像的细胞图像分割做遮罩。我做了阈值操作以将图像变成二进制蒙版，但是我需要将随机大小的中心图像放到36x36的蒙版中，并且我得到了大于此大小的图像。图像就是这样的。我想做的是以36x36的零矩阵为中心，但是我不习惯进行图像处理。

原始的是这个：

Answer 1

您好尝试此代码，我制作了一张图片供您理解：）

image = [[0, 0, 0, 0, 0],
         [0, 1, 0, 0, 0],
         [1, 1, 1, 0, 0], 
         [0, 1, 0, 0, 0],
         [0, 0, 0, 0, 0]]

image_width = 5
image_height = 5

lowest_x = -1
lowest_y = -1
bigest_x = -1
bigest_y = -1

# Get the square of the shape of your image (edge coordinate)
for y in range(len(image)):
    for x in range(len(image[y])):
        if image[y][x] != 0:
            if x < lowest_x or lowest_x == -1:
                lowest_x = x
            if y < lowest_y or lowest_y == -1:
                lowest_y = y
            if x > bigest_x or bigest_x == -1:
                bigest_x = x
            if y > bigest_y or bigest_y == -1:
                bigest_y = y

print ("Edge coordinate = " + str(lowest_y) + ":" + str(lowest_x) + "  -  " + str(bigest_y) + ":" + str(bigest_x))

chunk_width = bigest_x - lowest_x + 1
chunk_height = bigest_y - lowest_y + 1

print ("Chunk size = " + str(chunk_height) + " " + str(chunk_width))

y_delimiter = (image_height - chunk_height) / 2
x_delimiter = (image_width - chunk_width) / 2

print ("Start of new coord = " + str(y_delimiter) + " " + str(x_delimiter))

new_image = [[0 for i in range(image_height)] for j in range(image_width)]
for y in range(chunk_height):
    for x in range(chunk_width):
        new_image[y_delimiter + y][x + x_delimiter] = image[lowest_y + y][lowest_x + x]

print("")

for y in range(len(new_image)):
    print ' '.join(str(x) for x in new_image[y])

Answer 2

这是使用numpy 2D索引将一个图像插入另一个图像的一种方法。

Load the cell image as grayscale

Create a black image into which to recenter the cell data

Threshold the cell image using Otsu thresholding    

Get the contour(s) for the thresholded cell image

From each contour (presumably only one) get its bounding box and cut out the corresponding area of the gray image as roi

Compute the top left corner x and y offsets for centering the roi into the black image

Use numpy 2D array indexing to put the roi into the black image properly centered

输入：

import cv2
import numpy as np


# load image as grayscale
cell = cv2.imread('cell.png', cv2.IMREAD_GRAYSCALE)

# create 400x400 black image (larger than img) into which to  do the recentering
result = np.zeros((400,400), dtype=np.uint8)

# threshold input image with Otsu thresholding
ret, thresh = cv2.threshold(cell, 0, 255, cv2.THRESH_BINARY+cv2.THRESH_OTSU) 
cv2.imshow('THRESH', thresh)

# get contours --- presumably just one
contours = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
contours = contours[0] if len(contours) == 2 else contours[1]
for cntr in contours:
    x,y,w,h = cv2.boundingRect(cntr)
    print(x,y,w,h)
    roi=cell[y:y+h, x:x+w]
    # compute top left corner location to center roi in result image
    xoff = int((400 - w)/2)
    yoff = int((400 - h)/2)
    result[yoff:yoff+h, xoff:xoff+w] = roi
    # display result for each bounding box from contours
    cv2.imshow('CENTERED', result)

cv2.waitKey(0)
cv2.destroyAllWindows()

# save resulting centered image
cv2.imwrite('cell_centered.png', result)