How to choose pixels based on specific color

Question

I have such image as below and I want to change the pixel colors which have a specific BGR value of [99,30,233] or [255,31,101] to gray, and the rest to green.

Here is my code at the moment

image = cv.imread("path")
def change_img_color(image, image_dummy):
    image_dummy[np.logical_or(np.all(image==[99,30,233],axis=2), np.all(image==[255,31,101],axis=2))] = [95, 95, 95]
    image_dummy[np.logical_or(np.all(image==[146,61,65],axis=2),np.all(image==[147,177,218],axis=2))] = [95, 95, 95]

    image_dummy[np.logical_or(np.all(image==[54,67,244],axis=2),np.all(image==[180,187,42],axis=2))] = [0, 0, 0]

    image_dummy[np.logical_and(np.all(image_dummy!=[95,95,95],axis=2), np.all(image_dummy!=[0,0,0],axis=2))] = [50, 130, 110]

    return image_dummy

The problem is that when I want to transform the image to this form, logically it should be like this: but it ends up like this:

I can't figure out why that orange color does not change to green color, when I specified all the pixels that are not gray and are not black should be converted to green.

Answer 1

I think there's a typo somewhere and the red regions are not identified by the logic of your definition. To avoid such errors, it can be helpful to avoid "magic numbers".

Here's another approach to your problem based on a previous SO thread: https://stackoverflow.com/a/50215020/9870343.

image = cv2.imread("path")
cimage = image.copy()
color1 = np.array([99,30,233])
color2 = np.array([255,31,101])
gray = np.array([95,95,95])
green = np.array([110, 130, 50])
gscale = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)

mask1 = cv2.inRange(image, color1, color1)
mask2 = cv2.inRange(image, color2, color2)
mask = mask1 + mask2

image[gscale!=0] = green
image[mask!=0] = gray 

The output image: