Process Images with AWS Lambda and .NET Core

Generating less-than-full-sized images of user-uploaded content to an application such as a CMS, social platform, or blog is a fairly common process. Smaller image sizes enable better performance for your users and save on bandwidth costs when full resolution isn’t a necessity.

However, resizing an image to potentially multiple smaller sizes or doing other processing can cause your uploading/publishing user experience to suffer as it is often not an immediate response. If you are uploading images to a service such as Amazon S3, you can enlist the help of a Lambda Function to resize those images out of process, thus enhancing the upload experience.

AWS Lambda is a “serverless” service which allows you to run code without having to worry about provisioning or maintaining the servers behind the scenes. You can essentially run “any” code in a Lambda function, including Node.js, Python, .NET, and many more¹.

In this post, we will walk through how to create a Lambda function using .NET Core and C# to resize images that are uploaded to an S3 bucket. For purposes of this post, I will assume you have the .NET Core SDK installed, and are familiar with what an S3 bucket is and how to upload the images to the bucket.

Note: Lambda does not [yet] support .NET 5, so this post uses ASP.NET Core 3.1

Getting Started

To start, install the AWS .NET Lambda templates which include project templates to “bootstrap” creating Lambda functions in .NET. You can install these by running the following command in your favorite CLI:

$ dotnet new -i Amazon.Lambda.Templates

Note: for production uses, I likely would not use these templates as my image processor would be part of a larger application managed by CloudFormation.

If you run dotnet new in your CLI, you’ll notice several new Lambda-related project templates available. For this project, we will use the lambda.S3 template. Run the following command to create your project (within the directory you want your project to be created).

$ dotnet new lambda.S3

Once this is finished, open the project in VS Code (or your favorite editor) using code .. Your new project contains two folders, src, and test, each with a single project in them. Make sure to run dotnet restore to restore the Nuget packages.

You will notice a aws-lambda-tools-defaults.json file in the src/LambdaImage directory. This is a configuration file that has some default settings used with the .NET Lambda tools. If you have multiple AWS credential profiles on your machine, you’ll want to set the profile attribute to your desired profile.

We will cover Lambda deployment in a separate post

The Code

Enough introduction, let’s get into the code. To begin, we need to install an image processing library for Nuget. My personal favorite is Six Labors’ Image Sharp as it is easy to use and supports a variety of image types and processing functions.

To install the package, run the following command from your project directory:

$ dotnet add package SixLabors.ImageSharp

The FunctionHandler method takes an S3Event parameter that contains information about the S3 object we are being notified about (the image that was just uploaded). We will use this to get the actual object from S3 for processing.

In the try block, replace the templated code with:

// First, copy the original image out of the "upload" folder and into permanent storage
var originalSizeKey = GetResizedFileKey(s3Event.Object.Key);
await _s3Client.CopyObjectAsync(s3Event.Bucket.Name, s3Event.Object.Key, s3Event.Bucket.Name, originalSizeKey);

// Now, get the image data to use for resizing
byte[] imageBytes = null;
var contentType = "";

using (var objectResponse = await _s3Client.GetObjectAsync(s3Event.Bucket.Name, s3Event.Object.Key))
using (var ms = new MemoryStream())
{
    contentType = objectResponse.Headers.ContentType;
    await objectResponse.ResponseStream.CopyToAsync(ms);
    imageBytes = ms.ToArray();
}

I made a helper method to make renaming the files easier.

First, because our Lambda will be listening to events for specific key prefixes, we need to copy the image out of that “folder” and into our permanent storage (S3 doesn’t have true folders). Next, we need to retrieve the object by its key (“file path”) and save the content type and the bytes to use for processing.

Before we do that though, let’s create a simple object to represent our image sizes:

public class ImageSize
{
    public string Key { get; set; }
    public int Width { get; set; }
    public int Height { get; set; }

    public ImageSize(string key, int width, int height)
    {
        Key = key;
        Width = width;
        Height = height;
    }
}

Now, time for the resizing! Back in the try/catch block, add the following code:

var imageSizes = new List<ImageSize>
{
    new ImageSize("small", 400, 400),
    new ImageSize("medium", 1000, 1000),
    new ImageSize("large", 1600, 1600)
};

foreach (var imgSize in imageSizes)
{
    context.Logger.LogLine($"... Resizing to {imgSize.Key}");

    var resizedFileKey = GetResizedFileKey(s3Event.Object.Key, imgSize.Key);

    IImageFormat imageFormat;
    using (var image = Image.Load(imageBytes, out imageFormat))
    {
        using (var outStream = new MemoryStream())
        {
            image.Mutate(x => x.Resize(new ResizeOptions
            {
                Mode = ResizeMode.Max, // TODO: might want to make this a property on each image size
                Position = AnchorPositionMode.Center,
                Size = new Size(imgSize.Width, imgSize.Height)
            }));

            image.Save(outStream, imageFormat);

            var putObjectRequest = new PutObjectRequest
            {
                Key = resizedFileKey,
                BucketName = s3Event.Bucket.Name,
                ContentType = imageContentType,
                InputStream = outStream
            };
            await _s3Client.PutObjectAsync(putObjectRequest);
        }
    }

    context.Logger.LogLine($"... Resized and saved file '{s3Event.Object.Key}' to '{imgSize.Key}'");
}

// Delete the original
await _s3Client.DeleteObjectAsync(s3Event.Bucket.Name, s3Event.Object.Key);

return "Ok"; // Lambda expects a string to be returned to indicate success

Let’s parse through what our code is doing. For each image size in our list, we are generating a new filename using the naming helper. Then, using the byte array created earlier, load the image using ImageSharp and detect its format. Using the Mutate API, we are resizing the image to the dimensions defined in our ImageSize and save it to a new stream. Once that is complete, we can save it back to the bucket with its new key. Be sure to save it outside of your original “upload” folder so you don’t find yourself in a never-ending loop of resizing. Finally, delete the original image from the upload path.

Image name helper

This is a helper method I made to add the “size key” to the image filename:

private string GetResizedFileKey(string originalKey, string size = null)
{
    var filename = Path.GetFileName(originalKey);
    var newFileKeyPrefix = originalKey.Replace(OriginalKeyPrefix, "").Replace(filename, "");

    if (!string.IsNullOrEmpty(size))
    {
        filename = $"{size}-{filename}";
    }

    return $"{newFileKeyPrefix}{filename}";
}

Wrapping up

So there we have it. You’ve completed your Lambda function to resize images uploaded to an S3 bucket when it gets notified of objects being created. We still need to deploy the function to Lambda and set up the notifications on the bucket, which we’ll do in our next post.

Have any tips or comments for image processing or other S3-related lambda functions? Feel free to share them in the comments below.

PS: This post is a part of the 2020 C# Advent Calendar. Be sure to check it out!

PPS. Thanks to @q for proofing this post.

¹ https://aws.amazon.com/lambda/

Header photo by Maxwell Nelson on Unsplash

Process Images with AWS Lambda and .NET Core

December 18, 2020

Getting Started

The Code

Image name helper

Wrapping up