本文转自：/gongxijun/p/6117588.html
在图像的深度学习中，为了丰富图像训练集，更好的提取图像特征，泛化模型（防止模型过拟合），一般都会对数据图像进行数据增强，
数据增强，常用的方式，就是旋转图像，剪切图像，改变图像色差,扭曲图像特征，改变图像尺寸大小，增强图像噪音（一般使用高斯噪音，盐椒噪音）等.
但是需要注意，不要加入其他图像轮廓的噪音.
对于常用的图像的数据增强的实现，如下：
# -*- coding:utf-8 -*-"""数据增强1. 翻转变换 flip2. 随机修剪 random crop3. 色彩抖动 color jittering4. 平移变换 shift5. 尺度变换 scale6. 对比度变换 contrast7. 噪声扰动 noise8. 旋转变换/反射变换 Rotation/reflectionauthor: XiJun.Gongdate:-11-29"""from PIL import Image, ImageEnhance, ImageOps, ImageFileimport numpy as npimport randomimport threading, os, timeimport logginglogger = logging.getLogger(__name__)ImageFile.LOAD_TRUNCATED_IMAGES = Trueclass DataAugmentation:"""包含数据增强的八种方式"""def __init__(self):pass@staticmethoddef openImage(image):return Image.open(image, mode="r")@staticmethoddef randomRotation(image, mode=Image.BICUBIC):"""对图像进行随机任意角度(0~360度)旋转:param mode 邻近插值,双线性插值,双三次B样条插值(default):param image PIL的图像image:return: 旋转转之后的图像"""random_angle = np.random.randint(1, 360)return image.rotate(random_angle, mode)@staticmethoddef randomCrop(image):"""对图像随意剪切,考虑到图像大小范围(68,68),使用一个一个大于(36*36)的窗口进行截图:param image: PIL的图像image:return: 剪切之后的图像"""image_width = image.size[0]image_height = image.size[1]crop_win_size = np.random.randint(40, 68)random_region = ((image_width - crop_win_size) >> 1, (image_height - crop_win_size) >> 1, (image_width + crop_win_size) >> 1,(image_height + crop_win_size) >> 1)return image.crop(random_region)@staticmethoddef randomColor(image):"""对图像进行颜色抖动:param image: PIL的图像image:return: 有颜色色差的图像image"""random_factor = np.random.randint(0, 31) / 10. # 随机因子color_image = ImageEnhance.Color(image).enhance(random_factor) # 调整图像的饱和度random_factor = np.random.randint(10, 21) / 10. # 随机因子brightness_image = ImageEnhance.Brightness(color_image).enhance(random_factor) # 调整图像的亮度random_factor = np.random.randint(10, 21) / 10. # 随机因1子contrast_image = ImageEnhance.Contrast(brightness_image).enhance(random_factor) # 调整图像对比度random_factor = np.random.randint(0, 31) / 10. # 随机因子return ImageEnhance.Sharpness(contrast_image).enhance(random_factor) # 调整图像锐度@staticmethoddef randomGaussian(image, mean=0.2, sigma=0.3):"""对图像进行高斯噪声处理:param image::return:"""def gaussianNoisy(im, mean=0.2, sigma=0.3):"""对图像做高斯噪音处理:param im: 单通道图像:param mean: 偏移量:param sigma: 标准差:return:"""for _i in range(len(im)):im[_i] += random.gauss(mean, sigma)return im# 将图像转化成数组img = np.asarray(image)img.flags.writeable = True # 将数组改为读写模式width, height = img.shape[:2]img_r = gaussianNoisy(img[:, :, 0].flatten(), mean, sigma)img_g = gaussianNoisy(img[:, :, 1].flatten(), mean, sigma)img_b = gaussianNoisy(img[:, :, 2].flatten(), mean, sigma)img[:, :, 0] = img_r.reshape([width, height])img[:, :, 1] = img_g.reshape([width, height])img[:, :, 2] = img_b.reshape([width, height])return Image.fromarray(np.uint8(img))@staticmethoddef saveImage(image, path):image.save(path)def makeDir(path):try:if not os.path.exists(path):if not os.path.isfile(path):# os.mkdir(path)os.makedirs(path)return 0else:return 1except Exception, e:print str(e)return -2def imageOps(func_name, image, des_path, file_name, times=5):funcMap = {"randomRotation": DataAugmentation.randomRotation,"randomCrop": DataAugmentation.randomCrop,"randomColor": DataAugmentation.randomColor,"randomGaussian": DataAugmentation.randomGaussian}if funcMap.get(func_name) is None:logger.error("%s is not exist", func_name)return -1for _i in range(0, times, 1):new_image = funcMap[func_name](image)DataAugmentation.saveImage(new_image, os.path.join(des_path, func_name + str(_i) + file_name))opsList = {"randomRotation", "randomCrop", "randomColor", "randomGaussian"}def threadOPS(path, new_path):"""多线程处理事务:param src_path: 资源文件:param des_path: 目的地文件:return:"""if os.path.isdir(path):img_names = os.listdir(path)else:img_names = [path]for img_name in img_names:print img_nametmp_img_name = os.path.join(path, img_name)if os.path.isdir(tmp_img_name):if makeDir(os.path.join(new_path, img_name)) != -1:threadOPS(tmp_img_name, os.path.join(new_path, img_name))else:print 'create new dir failure'return -1# os.removedirs(tmp_img_name)elif tmp_img_name.split('.')[1] != "DS_Store":# 读取文件并进行操作image = DataAugmentation.openImage(tmp_img_name)threadImage = [0] * 5_index = 0for ops_name in opsList:threadImage[_index] = threading.Thread(target=imageOps,args=(ops_name, image, new_path, img_name,))threadImage[_index].start()_index += 1time.sleep(0.2)if __name__ == '__main__':threadOPS("/home/pic-image/train/12306train","/home/pic-image/train/12306train3")
更多数据增强方法：/aleju/imgaug
