img_processing.py

'''Supporting functions for image preprocessng, data loading and labeling.'''

import glob

import matplotlib.pyplot as plt
import natsort
import numpy as np
from PIL import Image
import pytesseract
from skimage.color import rgb2gray
from skimage.exposure import histogram
from skimage.filters import threshold_otsu
from skimage.io import imread
from skimage.transform import rescale
from skimage.util import invert, crop


def show_with_hist(img, title):
    '''Plot imgage alongside its histogram.'''
    plt.figure()
    plt.subplot(121)
    plt.imshow(img, cmap=plt.get_cmap('gray'))
    plt.title(title)
    hist, bins_center = histogram(img)
    plt.subplot(122)
    plt.plot(bins_center, hist, lw=2)


def crop_ui(img):
    '''Remove FLIR camera UI from image'''
    img_cropped = crop(img, ((27, 19), (25, 37)))
    return img_cropped


def get_temperature_bounds(img, bounds=(((6, 24), (283, 318)),
                                        ((219, 236), (283, 318)))):
    '''Extract temperature values from FLIR UI on image.'''
    img = invert(img)
    temp_txt = []
    for bound in bounds:
        bound_img = img[slice(*bound[0]), slice(*bound[1])]
        bound_img = rescale(bound_img, 4, anti_aliasing=True)
        thr = threshold_otsu(bound_img)
        img_txt = bound_img > thr
        img_txt = Image.fromarray(img_txt)
        temp = pytesseract.image_to_string(img_txt, config='digits')
        if temp != '':
            temp = float(temp) / 10
        else:
            temp = 0
        temp_txt.append(temp)
    return temp_txt


def full_prepare(img):
    '''Pipeline for FLIR images, convert to grayscale, crop ui and invert.'''
    img_gray = rgb2gray(img)
    img_crop = crop_ui(img_gray)
    img_prep = invert(img_crop)
    return img_prep


def load_img_series(path):
    '''
    Load jpg images containing glob pattern in path and get them in array.
    '''
    imgs = glob.glob(path + '*.jpg')
    imgs = natsort.natsorted(imgs)
    return [imread(img) for img in imgs]


def default_img_set():
    '''
    Get default set of metal grains cooling down recorded with FLIR
    thermovision camera.
    '''
    samples_names = ('104_E5R', '113_E5R', '119_E5R',
                     '107_E6R', '108_E6R', '117_E6R',
                     '105_E11R', '106_E11R', '115_E11R',
                     '111_E16R', '112_E16R', '118_E16R')

    X = [load_img_series('img/' + name) for name in samples_names]
    labels = (name.split('_', 1)[1] for name in samples_names)
    y = encode_labels(labels)
    return [X, y]


def decode_labels(y):
    '''Turn numeric labels into grain samples names.'''
    y_labels = np.array(['E5R', 'E6R', 'E11R', 'E16R'])
    return y_labels[y]


def encode_labels(y_labels):
    '''Turn grain samples names into numeric labels.'''
    y = {'E5R': 0, 'E6R': 1, 'E11R': 2, 'E16R': 3}
    return [y[label] for label in y_labels]