Store images in Amazon S3 behind Amazon CloudFront and use S3 Object Lambda to apply transformations to objects during GET requests is the simplest managed pattern for dynamic image processing and global delivery.

S3 Object Lambda lets you run Lambda code inline with S3 GET operations to resize or filter images on demand without creating derivative files or managing proxy infrastructure. When you pair S3 Object Lambda with CloudFront you get low latency and automatic scaling with minimal operational effort which matches the team preferences for a fully managed, highly available solution.

Store images in Amazon S3 and front them with Amazon CloudFront, using AWS Lambda@Edge to modify origin responses and dynamically resize the images is workable but not the easiest. Lambda@Edge has packaging and runtime constraints and it adds deployment complexity across Regions which makes it less straightforward than using S3 Object Lambda for per object transformations.

Store images in Amazon S3 behind Amazon CloudFront and configure S3 Event Notifications to invoke an AWS Lambda function to transform images when users request them is incorrect because S3 Event Notifications do not fire on GET requests. Notifications trigger on object creation and other lifecycle events which means you cannot perform transformations per retrieval with that pattern.

Store images in Amazon DynamoDB with Global Tables and invoke AWS Lambda to process items as they are read from the table is not appropriate because DynamoDB is not designed to serve or transform binary image data. That approach adds unnecessary complexity and does not provide the caching and global delivery benefits of S3 plus CloudFront.

Full AWS Practitioner Certification Question