Chapter 21 Working With Images

Images are a valueable potential source of information, and image processing, including image recognition, is also an important application of machine learning.

We assume you have loaded the following packages:

import numpy as np
import pandas as pd

## /home/otoomet/R/x86_64-pc-linux-gnu-library/4.5/reticulate/python/rpytools/loader.py:120: UserWarning: Pandas requires version '1.3.6' or newer of 'bottleneck' (version '1.3.5' currently installed).
##   return _find_and_load(name, import_)

import matplotlib.pyplot as plt

21.1 Loading images

Images can be read using matplotlib’s plt.imread(). This is a handy function that can read jpg and png images into an array. Below and example with a 4-layer color image. I’ll load a tiny image (\(8\times 5\) pixels) of Scottish flag:

flag = plt.imread("img/flag-of-scotland-5x8.png")
flag.shape

## (5, 8, 4)

np.set_printoptions(precision=2)
flag[:,:,0]  # R

## array([[0.93, 0.81, 0.31, 0.  , 0.  , 0.23, 0.77, 0.95],
##        [0.18, 0.74, 0.95, 0.64, 0.58, 0.95, 0.78, 0.25],
##        [0.  , 0.  , 0.55, 0.99, 1.  , 0.63, 0.  , 0.  ],
##        [0.04, 0.68, 0.96, 0.71, 0.66, 0.96, 0.73, 0.16],
##        [0.9 , 0.86, 0.38, 0.  , 0.  , 0.33, 0.82, 0.93]], dtype=float32)

flag[:,:,1]  # G

## array([[0.95, 0.84, 0.47, 0.37, 0.37, 0.43, 0.81, 0.96],
##        [0.41, 0.78, 0.96, 0.71, 0.66, 0.96, 0.82, 0.44],
##        [0.37, 0.37, 0.63, 1.  , 1.  , 0.69, 0.37, 0.37],
##        [0.38, 0.73, 0.96, 0.75, 0.72, 0.96, 0.77, 0.4 ],
##        [0.92, 0.88, 0.51, 0.37, 0.37, 0.48, 0.85, 0.94]], dtype=float32)

flag[:,:,2]  # B

## array([[0.97, 0.91, 0.75, 0.72, 0.72, 0.74, 0.9 , 0.98],
##        [0.73, 0.88, 0.98, 0.85, 0.83, 0.98, 0.9 , 0.74],
##        [0.72, 0.72, 0.82, 1.  , 1.  , 0.84, 0.72, 0.72],
##        [0.73, 0.86, 0.98, 0.87, 0.85, 0.98, 0.88, 0.73],
##        [0.95, 0.94, 0.77, 0.72, 0.72, 0.76, 0.92, 0.96]], dtype=float32)

_ = plt.imshow(flag)

plot of chunk unnamed-chunk-2

The dimension of the resulting array is \(5 \times 8 \times 4\). The first two dimensions, \(5\times 8\) denote the height and width of the image. imread() uses the common way of denoting the dimension of matrices–rows first (counting from top), and columns second. This may cause quite a headache as in case of pictures we normally think in terms of horizontal position first and vertical second (counting from bottom).