We’re developing novel theories and algorithms for humans and machines to capture, process, analyse and understand single and multiple images and video.
Image processing involves the development of novel theories and algorithms to capture, process, analysis and understand the images. This important and challenging research brings together experts from a range of fields to address fundamental signal-processing issues for which the input is an image, a series of images or a video. The most widely used image-processing application is Adobe Photoshop, and important applications can also be found in medicine, remote sensing and robot vision.
Industry partners: UBTech, Toyota and Huawei
Conventional image processing depends on domain knowledge to analyse images, identify filters and develop transformation, and pre-processing. Beyond this, modern image processing powered by deep learning, with the abundance of high-performance computing resources and big data, is gathering steam.
This project systematically studies deep-learning-based image processing, from the establishment of fundamental deep-learning theories to understanding the parameter structure of deep networks. We design network optimisation techniques, develop generic deep learning components for image-processing tasks (for example, image deblur and haze removing), and construct elegant, deep-network architectures for high-level image analysis and understanding (such as image classification, face recognition, and human pose recognition).
More than 120 papers have been published in prestigious journals and prominent conferences related to image processing, such as the Institute of Electrical and Eclectronics Engineers Transactions on Pattern Analysis and Machine Intelligence; Transactions on Image Processing; Computer Vision and Image Understanding; Conference on Computer Vision and Pattern Recognition; European Conference on Computer Vision; and the International Conference on Computer Vision.
Our technical achievements have been examined through practical challenges; for example, our significant successes in the ImageNet Large Scale Visual Recognition Challenge 2015 and 2017 demonstrated our technical expertise in this industry. This challenge, known as the ‘world cup’ in computer vision and machine learning, is where the latest technologies in image understanding are demonstrated and promoted, and the competition typically attracts strong attention from industry.