Image processing Backend

Technology stack

I used Spring Boot with Java 17 and Maven 3.6.3 for the backend. As database, I used MongoDB 4.4.4.

Branches

Each branch of {notify, forkJoinPool, javaUtilExecutor anc completableFuture} is a different approach for the backend multithreaded structure.

Endpoints

• api/uploadImage - used for uploading images
• api/downloadImage/{id} - for downloading images after an filter has been applied

Database

• I used MongoDB as a database, as it is a document oriented NOSQL database.
• Configuration and implementation can be found in repository package.

Services

• Located in package service.
• The actual image processing takes place in ImgSrv, as this class, not only creates the necessary additional data for multithreaded processing, but also has a nested class that implements Thread, called SubImageFilter.
• SubImageFilter calls the correct Filter class, as specified in process method of parent class.
• Each thread is given a slice of pixels rows from the image, and it operates

Filters

• Located in filter package.
• Each filter implementation is created using Filter Factory and has an applyFilter method, as specified by the Filter interface.
• Two filters are implemented, Black-and-white and sepia filters and the other are dummies, to show the value of the FilterFactory.
• Double-threshold may need fine-tuning of thresholds, none done yet.

Model

• The model class as well as DTO's are located in the model package.
• ImgBin is the model class, specifying the id of the image and the Binary representation.
• A successful upload of an image triggers the return of an answer withe ImageUploadResponse in the http body.

Synchronization

• I used a mutex for gradient and double threshold filters.
• The service ImgSrv, canny-edge-detection, double threshold filter and gradient filter use the barrier.

Internal representation

• In the utils package the classes used for the internal representation of the image are located along those used to encapsulate thread specific data, called ThreadSpecificData and ThreadSpecificDataT.
• Class Image saved the image in a matrix of pixels, where each pixel is represented by the Pixel class.
• At Image creation, a border is automatically added, as some filters need a border for their 3x3 kernels.
• The Pixel class saves internally the red, green, blue and transparency or alpha.
• Both the application.properties and the log4j2.xml file contain settings regarding the logging, the application.properties
  contains the logging level(without this setting, nothing gets printed) and the other the log4j2 settings.

Tests

• the test suite is a combination of unit tests and tests made to measure performance or simple tests. This project was created by Mihai Gheoace.