UMEC(HK) Publishes Research Paper on Deep Learning-based Salient Object Detection at 2020 IS&T International Symposium on Electronic Imaging


The Electronic Imaging Symposium has been held for 31 years, gathering the academia and industry leaders to share the latest updates about the imaging science and digital technologies. The 2020 IS&T International Symposium on Electronic Imaging (EI2020) took place in California, USA from 26 to 30 January 2020. UMEC(HK) had published a research paper on deep learning-based salient object detection. Meanwhile, Dr. Yuzhong Jiao, Senior Manager of UMEC(HK) was invited as a speaker of the symposium and to share the new research findings with the participants.

As a major breakthrough in artificial intelligence (AI), deep learning is increasing popular and powerful for promoting the development of 3D image/video processing and stereoscopic display. UMEC(HK)’s latest research paper titled “Dynamic Zero-Parallax-Setting Techniques for Multi-View Autostereoscopic Display” investigates how deep learning-based salient object detection is utilized to control zero plane setting for autostereoscopic display. The deep learning-based salient object detection methods can highlight most of the salient objects by computing the spatial distribution and moving the spatial centre of the salient objects to the zero-parallax plane, thus alleviating the blurry 3D vision. The paper compares two dynamic ZPS techniques, the first one is the maximum saliency computation by Graphic-Based Visual Saliency (GBVS) algorithm and the second one is the spatial distribution of salient objects by a convolutional neural networks (CNN)-based model. It was proven that both techniques help improve the 3D effect of autostereoscopic displays, while the second technique can achieve an even better 3D performance.

The research findings of UMEC(HK) have enhanced the autostereoscopic 3D display technology with the aid of deep learning. With the joint effort of the AI experts in UMEC(HK), more researches covering different applications is underway and will make positive impacts to every aspect of industries and lives.


Dr. Yuzhong Jiao, Senior Manager of UMEC(HK) shared the research findings on deep learning-based salient object detection at EI2020