augmentation.py

# ************************************************************
# Author : Bumsoo Kim, 2017
# Github : https://github.com/meliketoy/cellnet.pytorch
#
# Korea University, Data-Mining Lab
# Deep Convolutional Network Preprocessing Implementation
#
# Module : 1_preprocessor
# Description : augmentation.py
# The code for image augmentation.
# ***********************************************************

import cv2
import os
import numpy as np
import random

# Give random contrast to the image.
def random_contrast(image, lower=0.2, upper=1.8, seed=None):
    contrast_factor = np.random.uniform(lower, upper)
    return (image-np.mean(image))*contrast_factor + np.mean(image)

# Give random brightness emphasis to the image.
def random_brightness(image, max_delta=63, seed=None):
    delta = np.random.randint(-max_delta, max_delta)
    return image-delta

# Random Crop a portion of the image.
def random_crop(image, dim):
    if len(image.shape):
        W, H, D = image.shape
        w, h, d = dim
    else:
        W, H = image.shape
        w, h = size

    left, top = np.random.randint(W-w+1), np.random.randint(H-h+1)
    return image[left:left+w, top:top+h]

# Rotate the image.
def rotation(image, num, mode='random'):
    if(mode == 'strict'):
        deg = 90 * num
        (h,w) = image.shape[:2]
        center = (w/2, h/2)

        M = cv2.getRotationMatrix2D(center, deg, 1.0)
        rotated = cv2.warpAffine(image, M, (w,h))
    elif(mode == 'random'):
        deg = random.randrange(1, 360)
        (h,w) = image.shape[:2]
        center = (w/2, h/2)
        mean_val = [0.0, 0.0, 0.0]

        for channel in range(3):
            one_channel = image[:,:,channel]
            outer = np.append(one_channel[0,:-1], one_channel[-1,1:])
            tmp = np.append(one_channel[:-1,-1], one_channel[1:,0])
            outer = np.append(outer,tmp)
            outer = outer.flatten()
            mean_val[channel] = np.mean(outer[outer!=0]) # Fill the void area around the rotated image with the mean-value of the non-zero pixels

        M = cv2.getRotationMatrix2D(center, deg, 1.0)
        rotated = cv2.warpAffine(image, M, (w,h), borderMode=cv2.BORDER_CONSTANT, borderValue=mean_val)

    return rotated