[PYTHON] Extract image color (RGB)

Introduction

I had the opportunity to extract the colors (RGB) used from the image, so this is a memo.

environment

Language, package

image

lena.jpg

Check the size of the image

Check the image you want to use.

import cv2
import numpy as np

bgr_array = cv2.imread('lena.jpg')
print(bgr_array.shape)
# (512, 512, 3)

When loading images using OpenCV, the colors are in the order ** BGR **. Therefore, the image is that three 512 x 512 ** B **, ** G **, ** R ** matrices are lined up.

Get GBR

For example, get the BGR value at the top left below.

print(bgr_array[0, 0, :])
# [128 138 225]

Image resizing

With the above method, it is necessary to acquire the BGR value 512 x 512 times. Create the following matrix for efficient operation.

\left[
    \begin{array}{rrr}
      b_{0} & g_{0} & r_{0} \\
      \vdots & \vdots &  \vdots \\
      b_{n} & g_{n} & r_{n}
    \end{array}
  \right]

Where n = 512 * 512-1. Then, we will extract unique values in the row direction.

reshaped_bgr_array = bgr_array.reshape(512*512, 3)

# axis=Specify row uniqueness with 0
unique_bgr_array = np.unique(reshaped_bgr_array, axis=0)

#List of unique bgr values
print(unique_bgr_array)
"""
[[ 29  14  76]
 [ 31  11  86]
 [ 31  27 146]
 ...
 [224 175 191]
 [224 180 197]
 [225 247 253]]
"""

#Number of unique bgr values
print(len(unique_bgr_array))
# 73852

#Percentage of unique bgr values to total pixels
print(len(unique_bgr_array)/(512*512))
# 0.2817230224609375

Functionalization

The above code depends on the size of the image, so make it a function so that it can be reused.

def extract_unique_color(img_path, rgb=False):
    bgr_array = cv2.imread(img_path)
    row, col, _ = bgr_array.shape
    reshaped_bgr_array = bgr_array.reshape(row * col, 3)
    unique_color_array = np.unique(reshaped_bgr_array, axis=0)

    if rgb:
        #Sort elements into rgb
        unique_color_array = unique_color_array[:, [2, 1, 0]]

    return unique_color_array

Finally

I introduced how to extract unique colors in images.