在 scikit 中执行 Felzenszwalb 分割后，如何从图像分割转移到对 ROI 执行操作？

Question

In Python, I am using a background removal tool and scikit's Felzenszwalb algorithm to segment my image via

from rembg.bg import remove
import numpy as np
import io
from PIL import Image
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True

input_path = 'in.jpg'
output_path = 'out.png'

f = np.fromfile(input_path)
result = remove(f)
img = Image.open(io.BytesIO(result)).convert("RGBA")

#importing needed libraries
import skimage.segmentation
from matplotlib import pyplot as plt

#performing segmentation
res1 = skimage.segmentation.felzenszwalb(img, scale=500)

##uncomment to print results if desired
#fig = plt.figure(figsize=(12, 5))
#ax1 = fig.add_subplot(121)
#ax1.imshow(res1); ax1.set_xlabel("k=500")
#fig.suptitle("Graph based image segmentation")
#plt.tight_layout()

which gives me a plot of the segmented image in a few different colors. The output res1 is an ndarray "integer mask indicating segment labels" (from the scikit website). I now want to perform some color analysis on each of these regions. How would I use res1 to access each mask in turn (with either a homogeneous or transparent background) and perform my color analysis?

Answer 1

As you are using Numpy, you can access it with np.where

Among the regions you obtained with res1, choose the one you want to analyse (you can list the regions with np.unique(res1) ). For example for mask '2', do:

masked_img_2 = np.where(res1==2,img,0)
plt.imshow(masked_img_2) # if you want to visualize the result with matplotlib

this means: "keep img where res1 is equal to '2', elsewhere set it to Zero" https://numpy.org/doc/stable/reference/generated/numpy.where.html