Faculty of Computing and Informatics,
Multimedia University (MMU), Malaysia
Dr. Wong is the Deputy Dean of Research and Innovation at Faculty of Computing and Informatics, Multimedia University, Malaysia. Lai Kuan received her B.Sc. degree in Computer Science from Universiti Sains Malaysia (USM), and M.Sc. and Ph.D. degrees in Computing from the School of Computing, National University of Singapore (NUS). She was a recipient of the NUS research scholarship. Her research interests include computational photography, automatic artistic creations using interactive graphics and image manipulation, stereo image and video enhancement, 3D visualization and medical imaging.
Lai Kuan has published in several reputable conferences including ACM Multimedia, IEEE International Conference on Image Processing, IEEE Winter Conference on Computer Vision and Computer Graphics International, and journals such as IEEE Transactions on Multimedia and Multimedia Tools & Application. Notably, her research work won her several gold awards in various exhibitions such as ITEX, PECIPTA and PERINTIS from 2016 to 2018. For her dedication in teaching, she was also awarded the Excellent Teaching Award by Multimedia University in 2014.
Lai Kuan is the Technical Committee of APSIPA Signal Processing System. She was involved in organizing several international conferences/workshops. She was the Co-Chair for the International Workshop on Learning Semantics for Multimedia Big Data @ACPR 2015 and Workshop on AI Aesthetics in Art and Media @ ACCV 2018 as well as the organizing committee for Pacific-Rim Conference on Multimedia 2014, International Workshop in Advanced Image Technology 2018, and IEEE Multimedia Signal Processing 2019. In addition, she has served as the Technical Program Committee for conferences such as ACM Multimedia and Multimedia Modeling, and the Reviewer for several prominent journals including IEEE Transactions on Multimedia, IEEE Transactions on Image Processing, IEEE Transactions on Circuits and Systems for Video Technology, and International Journal on Computer Vision.
Keynote Title: Aesthetics-driven Image Recomposition: From Classic to Deep Vision
With the rapid advancement and adoption of mobile and digital imaging technology, almost everyone is now a photographer. According to Agence France Presse (AFP) report, 880 billion photographs were taken in 2014, and 27,800 photos/minute and 208,300 photos/minute, are uploaded to Instagram, and Facebook respectively. With the large amount of photos being captured and shared on social media, there is a growing interest in aesthetics quality improvement. One aspect of photography that contributes to high quality photos is image composition; the spatial arrangement of photo subjects in the image frame. Professional photographers often apply a wealth of photographic composition rules, i.e., rule of thirds, visual balance and simplicity to capture compelling photos. However, casual photographers may not have the knowledge of taking good photos and often find their photos unsatisfying. This fosters the need for aesthetics-driven recomposition tools that can assist them to enhance the aesthetics of their poorly-taken photos with ease.
Earlier works of image recomposition aim to produce resulting images that mimic professional photographs by employing optimization-based techniques that are guided by selected photographic composition techniques / rules. Moving forward, researchers then adopt the learning-based approach to learn an image recomposition model using traditional machine learning algorithms. More recently, with the introduction of generative networks, image recomposition can now be performed in an end-to-end manner, whereby an aesthetics prediction model acts as a discriminator to guide a generator network to produce resulting images with higher aesthetics appeal. This tutorial commence with the formulation of the image recomposition problem, followed by the chronological development of the image recomposition research; from the classic approaches that utilize optimization and traditional machine learning to deep vision approaches that employ convolutional neural network and generative network. The tutorial will conclude with the discussion on the challenges and opportunities in aesthetics-driven image recomposition.