Image Enhancement Platform

Client: Global Medical Devices Company 

Industry: Health

Goal: Design a system that takes low-quality, low-resolution images and improves them to look like conventional high-resolution images.

Tech: LessRay, Mask R-CNN, Python, Neural Network (NN), Tensorflow

 

getDeveloperSpeak to us about your software development requirements: contact@getdeveloper.net

 

THE CHALLENGE

• Embed a neural network which isolates specific areas on the low-resolution image, and also to grasp how a neural network works and how it ties together with Mask R-CNN framework.

• Design an algorithm that would locate areas on a full-resolution image, and thereby replace low-resolution, grainy areas from the appropriate high-resolution image using geometric transformations.

THE APPROACH

Collaborate with the client on a proof of concept project which would completely change the way that image enhancing platforms would operate, while also grasping how the current system operates, and the logic behind X-ray imaging. Instead of just enhancing low-resolution images that an X-ray device records, it would integrate a neural network into the system to further manipulate the image beyond any other image enhancing platform.

THE SOLUTION

• We manipulated the neural network with object segmentation, and trained it using a vast amount of images that the client provided for NN calibration. We used fresh images to test how well the recognition algorithm worked.

• We set up the network and framework running in the environment and successfully used Python for zone extraction.

THE RESULTS

An image enhancing tool that transforms low quality images and improves them. After the framework locates and marks the areas of interest (zones) on the low-resolution image, the same zones would be located on the high-resolution image and replaced accordingly.

 

getDeveloperSpeak to us about your software development requirements: contact@getdeveloper.net