# Image-processing filters using decorator pattern in Python 3

The goal of the following short Python 3 program is to improve my understanding of the decorator design pattern. Therefore, I would be mostly interested about feedback on this aspect, i.e. whether this is a correct usage of mentioned pattern. Also, I would like some comments regarding Python coding best practices, i.e. if there are ways to make my code more "Pythonesque". Of course, any other hints and remarks (e.g., related to the chosen use case, image editing), are welcome too.

#!/usr/bin/python3

import PIL.Image, PIL.ImageFilter, PIL.ImageOps

class IImage:
def get_img_obj(self):
raise NotImplementedError("cannot call method of IImage.getImgArr on the interface")

class BaseImage(IImage):
def __init__(self, img_obj):
self.img_obj = img_obj

def get_img_obj(self):
return self.img_obj

class ImageDecorator(IImage):
def get_img_obj(self):
raise NotImplementedError("cannot call method of ImageDecorator.getImgArr on the base class")
def __init__(self, img):
self.img = img

class BlurFilter(ImageDecorator):
def __init__(self, img, radius):
ImageDecorator.__init__(self, img)
def get_img_obj(self):
imageIternal = self.img.get_img_obj()

class ImageFlipper(ImageDecorator):
def __init__(self, img):
ImageDecorator.__init__(self, img)
def get_img_obj(self):
imageIternal = self.img.get_img_obj()
return PIL.ImageOps.flip(imageIternal)

im_raw = PIL.Image.open("lena.gif")
im_raw = im_raw.convert('L')
im_obj = BaseImage(im_raw)
im_obj_flipped = ImageFlipper(im_obj)
im_obj_flipped_blurred = BlurFilter(im_obj_flipped, 5)

im_obj.get_img_obj().show("original")
im_obj_flipped.get_img_obj().show("flipped")
im_obj_flipped_blurred.get_img_obj().show("flipped and blurred")


First of all, since Python is dynamically typed, there is not really a need for interfaces. That’s why there is no such language concept. So your IImage type does not really serve any need.

Similarly, the base type ImageDecorator is not really that useful. In statically typed language you are using these design patterns, and ultimately such a base type, because you need to interact with some common base type. Since we have duck typing in Python, this is not really necessary.

Following the Zen of Python, “Simple is better than complex”; and looking at those classes you wrote, it looks somewhat complex to me considering what it does. Compare it with the following solution:

#!/usr/bin/python3
import PIL.Image, PIL.ImageFilter, PIL.ImageOps

def flip(image):
return PIL.ImageOps.flip(image)

im_obj = PIL.Image.open('lena.gif')
im_obj = im_raw.convert('L')
im_obj_flipped = flip(im_obj)
im_obj_flipped_blurred = blur(im_obj_flipped, 5)

im_obj.show('original')
im_obj_flipped.show('flipped')
im_obj_flipped_blurred.show('flipped and blurred')


I think this is a lot more to the point, and it shows exactly what’s going on without unnecessarily wrapping the image in an object that does not really serve any purpose.

Of course, I don’t know if your BaseImage is actually needed for something; but if it’s just what you’re showing here, then it’s just a wrapper that doesn’t do anything on its own. You should avoid creating types that don’t have any behavior, especially if they only contain a single object. It’s much easier to just use that actual object then instead.

That being said, if the BaseImage actually is more complicated, then you can still keep your filters simple:

def blur(image, radius):
rawImage = image.get_img_obj()