Convert BGR to RGB (Opencv)

Question

I have a code that finds the edges around what was labeled in the image and places the value of -1 in the pixels of that border found. I’m having trouble converting the image already with the edges -1 in its value, from BGR to RGB. because when trying to convert, the value -1 is lost and only find values 0 (rest) and 255 (labeled) in the saved image. I wanted to keep the values 0, -1, 255 when saving the image in RGB format.

Follows code:

import cv2
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import matplotlib
import urllib.request
from PIL import Image

def mostrar_imagem(img):
    plt.imshow(img)
    plt.colorbar()
    plt.show()

img = cv2.imread('framepng/frame24.png',0)

kernel = np.ones((5,5),np.uint8)
erosion = cv2.erode(img, kernel, iterations = 1)
img = cv2.dilate(erosion, kernel, iterations = 1)

_, binario = cv2.threshold(img, 90, 255, cv2.THRESH_BINARY)

copia = img.copy()

tmp = cv2.findContours(binario, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
contornos = tmp[0] if len(tmp) == 2 else tmp[1]

img_contornos = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
mostrar_imagem(img_contornos)

cv2.drawContours(img_contornos, contornos, -1, (0, 255, 0), 1)
mask = cv2.inRange(img_contornos, (0, 254, 0), (0, 255, 0))
mostrar_imagem(mask)

binario = np.int16(binario)
binario[mask == 255] = -1
binario = np.ma.masked_where(binario == -1, binario)

cmap = matplotlib.cm.Greys  # Can be any colormap that you want after the cm
cmap.set_bad(color='red')

plt.show()
matplotlib.image.imsave('framepng-1/teste3.png', binario)

Answer 1

Problem

With img = cv2.imread('framepng/frame24.png',0) the image is converted to grayscale, because the flag 0 is used. With this, you lose all the color information that was present in the image.

Follow function mode flags imread():

cv::ImreadModes {
  cv::IMREAD_UNCHANGED = -1,
  cv::IMREAD_GRAYSCALE = 0,
  cv::IMREAD_COLOR = 1,
  cv::IMREAD_ANYDEPTH = 2,
  cv::IMREAD_ANYCOLOR = 4,
  cv::IMREAD_LOAD_GDAL = 8,
  cv::IMREAD_REDUCED_GRAYSCALE_2 = 16,
  cv::IMREAD_REDUCED_COLOR_2 = 17,
  cv::IMREAD_REDUCED_GRAYSCALE_4 = 32,
  cv::IMREAD_REDUCED_COLOR_4 = 33,
  cv::IMREAD_REDUCED_GRAYSCALE_8 = 64,
  cv::IMREAD_REDUCED_COLOR_8 = 65,
  cv::IMREAD_IGNORE_ORIENTATION = 128
}

Solution

If your image is in BGR and you want to convert to RGB, just use the following Python function:

img = img[:,:,::-1]

Or to improve the understanding of the code:

img_bgr = img_rgb[:,:,::-1]

Because an image is a matrix, with intensity values at each pixel coordinate. In the case of the RGB color space, each pixel will have a vector with 3 color channels, that is, 3 values. Then ::-1 is used to reverse the order of these values.

Or conversion directly by Opencv function:

img_rgb= cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)

But I believe that this would not solve your problem the way the code is presented, what should be done is to read the image as colored img = cv2.imread('framepng/frame24.png',1) or the original image img = cv2.imread('framepng/frame24.png',-1). And then convert this image to grayscale with another variable name, for example, cinza = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY). So the color/original image is kept in memory and can be manipulated later.

That is, the color space used in the code is grayscale (1 channel) and not BGR (3 channels). Because the functions used, such as morphological operations (erode and dilate), among others, are used in grayscale and need an image in this color space as input. Then converting from BGR to RGB wouldn’t solve your problem, because that’s not the problem.

And when adding the value of -1 to the color image of 3 channels, it would not be possible to use the functions of Opencv, because its matrix would be hybrid with some pixels with 3 components and others with 1 component. Generating more errors in future processing in this image, what can be accomplished is the insertion of -1 in one of the channels or in the 3 channels, to maintain the image with 3 channels.

Read a little about the spaces of colors