How to resize the new uploaded images using PIL before saving?

image = image.resize(size, Image.ANTIALIAS)

resize is non-destructive, it returns a new image.

here is what worked for me, inspired a little from django-resized

@receiver(pre_save, sender=MyUser)
@receiver(pre_save, sender=Gruppo)
def ridimensiona_immagine(sender, instance=None, created=False, **kwargs):
    foto = instance.foto

    foto.file.seek(0)
    thumb = PIL.Image.open(foto.file)
    thumb.thumbnail((
        200, 
        200
        ), PIL.Image.ANTIALIAS)


    buffer = StringIO.StringIO()
    thumb.save(buffer, "PNG")
    image_file = InMemoryUploadedFile(buffer, None, 'test.png', 'image/png', buffer.len, None)

    instance.foto.file = image_file

I searched for a solution to resize uploaded photo before saving. There are a lot of info bit and bit here and there (in StackOverflow). Yet, no complete solution. Here is my final solution that I think works for people who wants it.

Development Highlight

• Using Pillow for image processing (two packages required: libjpeg-dev, zlib1g-dev)
• Using Model and ImageField as storage
• Using HTTP POST or PUT with multipart/form
• No need to save the file to disk manually.
• Create multiple resolutions and stores their dimensions.
• Did not modify the Model itself

Install Pillow

$ sudo apt-get install libjpeg-dev
$ sudo apt-get install zlib1g-dev
$ pip install -I Pillow

myapp/models.py

from django.db import models

class Post(models.Model):
    caption = models.CharField(max_length=100, default=None, blank=True)
    image_w = models.PositiveIntegerField(default=0)
    image_h = models.PositiveIntegerField(default=0)
    image = models.ImageField(upload_to='images/%Y/%m/%d/', default=None, 
                blank=True, width_field='image_w', height_field='image_h')
    thumbnail = models.ImageField(upload_to='images/%Y/%m/%d/', default=None,
                blank=True)

This model saves a photo with thumbnail and an optional caption. This should be similar to the real-world use case.

Our goal is to resize the photo to 640x640 and generate a 150x150 thumbnail. We also need to return the dimension of the photo in our web API. image_w and image_h is a cached dimension in table. Note that we need to add quota ' when we declare an ImageField. However, thumbnail does not need the width and height field (so you can see the different).

That's the model. We don't need to override or add any functions to the model.

myapp/serializers.py

We will use a ModelSerializer to process the incoming data from HTTP POST.

from rest_framework import serializers
from myapp.models import Post

class PostSerializer(serializers.ModelSerializer):
    class Meta:
        model = Post
        fields = ('caption',)

We do not want the serializer to handle the uploaded photo - we will do it ourself. So, no need to include 'image' in fields.

myapp/views.py

@api_view(['POST'])
@parser_classes((MultiPartParser,))
def handle_uploaded_image(request):
    # process images first.  if error, quit.
    if not 'uploaded_media' in request.FILES:
        return Response({'msg': 'Photo missing.'}, status.HTTP_400_BAD_REQUEST)
    try:
        im = Image.open(StringIO(request.FILES['uploaded_media'].read()))
    except IOError:
        return Response({'msg': 'Bad image.'}, status.HTTP_400_BAD_REQUEST)

    serializer = PostSerializer(data=request.DATA, files=request.FILES)
    if not serializer.is_valid():
        return Response({'msg': serializer.errors}, status.HTTP_400_BAD_REQUEST)

    post = Post.create()
    if serializer.data['caption'] is not None:
        post.caption = serializer.data['caption']

    filename = uuid.uuid4()
    name = '%s_0.jpg' % (filename)
    post.image.save(name=name, content=resize_image(im, 640))
    name = '%s_1.jpg' % (filename)
    post.thumbnail.save(name=name, content=resize_image(im, 150))

    post.save()
    return Response({'msg': 'success',
        'caption': post.caption,
        'image': {
            'url': request.build_absolute_uri(post.image.url),
            'width': post.image_w,
            'height': post.image_h,
        }
        'thumbnail': request.build_absolute_uri(post.thumbnail.url),
    }, status.HTTP_201_CREATED)

helper functions in a shared py

def resize_image(im, edge):
    (width, height) = im.size
    (width, height) = scale_dimension(w, h, long_edge=edge)
    content = StringIO()
    im.resize((width, height), Image.ANTIALIAS).save(fp=content, format='JPEG', dpi=[72, 72])
    return ContentFile(content.getvalue())

def scale_dimension(width, height, long_edge):
    if width > height:
        ratio = long_edge * 1. / width
    else:
        ratio = long_edge * 1. / height
    return int(width * ratio), int(height * ratio)

Here, we do not use Image.thumbnail because it will change the original image. This code is suitable if we want to store multiple resolutions (such as low res and high res for different purpose). I found that resizing the incoming message twice will degrade the quality.

We also do not save the image directly to disk. We push the image via a ContentFile (a File object for content in memory) to ImageField and let the ImageField do its job. We wants the multiple copies of image have the same file name with different postfix.

Credits

Here are the links of codes that I had references to build this solution:

• Use of ContentFile in ImageField.save, by Martey: https://stackoverflow.com/a/7022005/3731039
• Use of StringIO for reading multipart/form, by johndoevodka: https://stackoverflow.com/a/15523422/3731039
• Code for resizing: http://davedash.com/2009/02/21/resizing-image-on-upload-in-django/

If you want to validate the image as well, see this: https://stackoverflow.com/a/20762344/3731039

I use django-resized for my projects.