Python matplolib ListedColormap paints all image yellow but identical almost image paints correctly-CodePudding

Stumbled to one very strange problem.I create a tgi image from agro images.Althought i have two very simiral agro images with almost the same colours ,one of the tif files produces a good tgi image but the other produces a tgi image as full yellow.I apply a matplolib.ListedColormap where i think my code bugs:

import cv2
import numpy as np
from PIL import Image
from matplotlib import colors




HD_rgb_tgi =  cv2.imread('good_tif.tif',cv2.IMREAD_UNCHANGED)
red = HD_rgb_tgi[:,:,0].astype('float64')
green = HD_rgb_tgi[:,:,1].astype('float64')
blue = HD_rgb_tgi[:,:,2].astype('float64')
tgi=(-1) * 0.5 * ((200*(red-green))-(100 * (red-blue)))
tgi_normalized = ((tgi - np.amin(tgi)) / (np.amax(tgi) - np.amin(tgi)))
tgi_f = Image.fromarray(np.uint8(colors.ListedColormap([ '#C90000', '#CF630A'  , '#EBE42D', '#27FF0F', '#0C750C'])(tgi_normalized) * 255))
tgi_f = tgi_f.convert('RGBA')
tgi_f=cv2.cvtColor(np.float32(tgi_f), cv2.COLOR_RGBA2BGRA)
cv2.imwrite("tgi_good.png",tgi_f)





HD_rgb_tgi =  cv2.imread('bad_tif.tif',cv2.IMREAD_UNCHANGED)
red = HD_rgb_tgi[:,:,0].astype('float64')
green = HD_rgb_tgi[:,:,1].astype('float64')
blue = HD_rgb_tgi[:,:,2].astype('float64')
tgi=(-1) * 0.5 * ((200*(red-green))-(100 * (red-blue)))
tgi_normalized = ((tgi - np.amin(tgi)) / (np.amax(tgi) - np.amin(tgi)))
tgi_f = Image.fromarray(np.uint8(colors.ListedColormap([ '#C90000', '#CF630A'  , '#EBE42D', '#27FF0F','#0C750C'])(tgi_normalized) * 255))
tgi_f = tgi_f.convert('RGBA')
tgi_f=cv2.cvtColor(np.float32(tgi_f), cv2.COLOR_RGBA2BGRA)
cv2.imwrite("tgi_bad.png",tgi_f)

link to the 2 tif images:

https://mega.nz/file/2J0EzbKQ#wPFPr4onmRCwEvdOHyBCwLQnBMqPQYIsZucA8bNERBw https://mega.nz/file/DF0i2BTJ#HuzKxWI60-R0DjAB7lwuqkhOiO5Y8bgEFXUHkHfmwNE

Sorry for the external links but i can't reproduce the problem to other images. Best regards

CodePudding user response：

Even if not visible, apparently the histograms of the photos were not the same. Applying cv2.equalizeHist to the bad image seems to do the trick.

import cv2
import numpy as np
from PIL import Image
from matplotlib import colors


HD_rgb_tgi =  cv2.imread('good_tif.tif',cv2.IMREAD_UNCHANGED)#odm_orthophoto
red = HD_rgb_tgi[:,:,0].astype('float64')
green = HD_rgb_tgi[:,:,1].astype('float64')
blue = HD_rgb_tgi[:,:,2].astype('float64')
tgi=(-1) * 0.5 * ((200*(red-green))-(100 * (red-blue)))
tgi_normalized = ((tgi - np.amin(tgi)) / (np.amax(tgi) - np.amin(tgi)))
tgi_f = Image.fromarray(np.uint8(colors.ListedColormap([ '#C90000', '#CF630A'  , '#EBE42D', '#27FF0F', '#0C750C'])(tgi_normalized) * 255))
tgi_f = tgi_f.convert('RGBA')
tgi_f=cv2.cvtColor(np.float32(tgi_f), cv2.COLOR_RGBA2BGRA)
cv2.imwrite("tgi_good.png",tgi_f)


def run_histogram_equalization(image_path):
    rgb_img = cv2.imread(image_path)
    # convert from RGB color-space to YCrCb
    ycrcb_img = cv2.cvtColor(rgb_img, cv2.COLOR_BGR2YCrCb)
    # equalize the histogram of the Y channel
    ycrcb_img[:, :, 0] = cv2.equalizeHist(ycrcb_img[:, :, 0])
    # convert back to RGB color-space from YCrCb
    equalized_img = cv2.cvtColor(ycrcb_img, cv2.COLOR_YCrCb2BGR)
    cv2.imwrite('res.png',equalized_img)

run_histogram_equalization('bad_tif.tif')


D_rgb_tgi =  cv2.imread('res.png',cv2.IMREAD_UNCHANGED)#odm_orthophoto
red = HD_rgb_tgi[:,:,0].astype('float64')
green = HD_rgb_tgi[:,:,1].astype('float64')
blue = HD_rgb_tgi[:,:,2].astype('float64')
tgi=(-1) * 0.5 * ((200*(red-green))-(100 * (red-blue)))
tgi_normalized = ((tgi - np.amin(tgi)) / (np.amax(tgi) - np.amin(tgi)))
tgi_f = Image.fromarray(np.uint8(colors.ListedColormap([ '#C90000', '#CF630A'  , '#EBE42D', '#27FF0F', '#0C750C'])(tgi_normalized) * 255))
tgi_f = tgi_f.convert('RGBA')
tgi_f=cv2.cvtColor(np.float32(tgi_f), cv2.COLOR_RGBA2BGRA)
cv2.imwrite("tgi_bad.png",tgi_f)