HomeProjectsInformation and Communication TechnologiesAI Based Color Image Compression Platform

AI Based Color Image Compression Platform

Project Quick Facts

Principal Investigator

  • Prof. WONG Tien Tsin

    Department of Computer Science and Engineering

  • Funding

    Research Grants Council of Hong Kong

With the advancement of digital photography technology, the demand for cloud and local storage space is increasing. Image processing technology nowadays can convert color images into grayscale ones and greatly reduce the storage space required, however, original colour cannot be restored once an image is converted to grayscale. We therefore develop a color conversion system based on convolutional neural network. The system provides an innovative method to synthesise invertible grayscale, which offers color-to-gray conversion and grayscale-to-color restoration, to fully restore the original colour of grayscale images. It can be applied in image processing industry and cloud storage applications.

  • We propose a color-encoding learning model which consists of an encoding network to convert a color image to grayscale, and a decoding network to invert the grayscale to color
  • The color image recovered by the system is visually the same as the original image
  • Cloud and local storage space can be greatly reduced as the images are significant compressed using AI-based grayscale generation and color recovery technology


Online/Local Data Storage

Education purpose in drawing application

Target Users:

Cloud Storage Service Provider

Hard Disk Manufactures

Educational Professionals

Given an input color image (left), our method converts it to an invertible grayscale (middle) that can be later restored to the color version (right). The key of our method is to encode the original color information in the generated grayscale as unobvious pattern (blown-ups of the middle image) via a convolutional neural network.

Do you like our project?