VSPC Rising Stars Talks
IEEE VSPC-TC is pleased to announce the finalists of the IEEE VSPC Rising Star Award. Learn more about the finalists and their research contributions below.
Biography
Hadi Amirpour (IEEE Senior Member) is a Tenure-Track Assistant Professor at the University of Klagenfurt, Austria, where he received his Ph.D. in Computer Science in 2022. His research focuses on efficient and perceptually aware visual communication, spanning content-aware and machine-learning-based video encoding, video complexity analysis, adaptive and live streaming, visual quality assessment,......
Hadi Amirpour (IEEE Senior Member) is a Tenure-Track Assistant Professor at the University of Klagenfurt, Austria, where he received his Ph.D. in Computer Science in 2022. His research focuses on efficient and perceptually aware visual communication, spanning content-aware and machine-learning-based video encoding, video complexity analysis, adaptive and live streaming, visual quality assessment, and immersive media. He has contributed to JPEG Pleno standardization and is an inventor on multiple patents and patent applications in video encoding, streaming, and visual media processing.
His research has received several prestigious awards, including Best Paper Awards at MMSP 2026, NAB 2025, and PCS 2024, as well as the Best New Streaming Innovation Award. His work was also recognized as a Best Paper Candidate at QoMEX 2026 and VCIP 2024. In addition, he secured first place in two VCIP Grand Challenges focused on live broadcasting video quality assessment and image manipulation quality assessment.
He contributes to the research community as an Associate Editor of ACM TOMM and IEEE TCSVT, and as a member of the IEEE VSPC TC and EURASIP TAC VIP. His academic service also includes guest-editing special issues of Pattern Recognition and the Journal of Visual Communication and Image Representation, as well as organizing special sessions, workshops, and tutorials. His conference leadership roles include Lead TPC Chair for VCIP 2025 and General Co-Chair for QoMEX 2026, with an upcoming role as General Co-Chair for IEEE MMSP 2027.
Abstract
Newer codecs promise the same visual quality at lower bitrates, but their heavy computational demands make them hard to deploy in latency-sensitive live streaming. Until that gap closes, workhorse standards like AVC will keep powering much of the world’s live video. So what if the next leap in live streaming came from making the codecs we already deploy content-aware?
The talk first shows how content-adaptive encoding unlocks hidden potential in existing codecs. Every video is different, and a static news anchor should not be encoded the same way as a fast-moving football match. By finding the optimal encoding parameters for each piece of content, the same codec can deliver higher visual quality at lower bitrates. The catch is that finding these parameters through brute-force search is far too slow for live delivery.
This is where video complexity analysis comes in. Lightweight features that capture a video’s spatial and temporal characteristics can be extracted in real time and used to predict effective encoding settings, bringing the gains of content-adaptive encoding to live streaming within tight bitrate and latency budgets. The talk walks through several approaches to complexity analysis, from early prototypes to deployable techniques.
Smarter encoding is only half the story; live services also need to know how good their streams actually look. Conventional quality metrics are too slow and costly to run in real time, so the talk then shows how information already generated during encoding can be repurposed to estimate compressed-video quality on the fly, at almost no extra cost.
Looking ahead, the vision is to carry this content awareness beyond 2D video into 3D and immersive media, such as volumetric video and 3DGS, where data volumes are far larger and every bit saved matters even more.
Biography
Zhengxue Cheng is currently an Assistant Professor at the School of Information Science and Electronic Engineering, Shanghai Jiao Tong University. She received her B.E. degree from Shanghai Jiao Tong University, Shanghai, China, in 2014, and her M.E. degrees from Waseda University, Japan, and Shanghai Jiao Tong University, China, in 2015 and 2017, respectively, through a double-degree program.......
Zhengxue Cheng is currently an Assistant Professor at the School of Information Science and Electronic Engineering, Shanghai Jiao Tong University. She received her B.E. degree from Shanghai Jiao Tong University, Shanghai, China, in 2014, and her M.E. degrees from Waseda University, Japan, and Shanghai Jiao Tong University, China, in 2015 and 2017, respectively, through a double-degree program. From October 2018 to February 2019, she was a visiting intern at EPFL, Switzerland. She received her Ph.D. degree from Waseda University in 2020. She was awarded the Research Fellowship for Young Scientists from the Japan Society for the Promotion of Science (JSPS). From 2020 to 2024, she was an algorithm expert at Ant Group (Alipay), China.
Her research interests include multimedia data compression and processing, with a particular focus on learning-based image and video coding. She has published over 100 journal and conference papers, primarily in leading venues such as TCSVT, TIP, TMM, TBC, VCIP, ISCAS, CVPR, ICLR, NeurIPS, ACM MM, AAAI, ICME, ICASSP, ICIP, and EMNLP. Her most representative work (Cheng2020) introduces an end-to-end learned image compression method that integrates Gaussian Mixture Models with lightweight attention modules. It was the first learned codec in 2020 to achieve compression performance on par with the VVC/H.266 all-intra profile. As of October 2026, this paper has received over 1,830 citations and has served as a foundational baseline for many subsequent end-to-end image and video codecs.
She has received 6 Best Paper Awards, most notably at IEEE VCIP 2025, IEEE VCIP 2024, the UHD World Association 2030 Outstanding Paper Award, IEEE BMSB 2026, and the TeleCom Technical Paper Awards in 2020 and 2023. She has also received 4 Challenge Awards, including the Winner Award of the ACM MM 2026 Egolink Challenge, the Winner Award of the CVPR NTIRE 2025 Efficient Super-Resolution Challenge, the Runner-up Award of the CVPR NTIRE 2025 Diffusion-based SR Challenge, and the Silver Award at the PCS 2019 Grand Challenge on Video Coding.
In terms of academic service, she currently serves as an Associate Editor for IEEE TCSVT and APSIPA TSIP. She has delivered two tutorials, both related to "Multi-modal Data Compression," for ISCAS 2026 and VCIP 2026. She was the co-organizer of a special session on "Advanced Representation Models for Image and Video Compression" at PCS 2025. In addition, she has regularly served as a reviewer for numerous top-tier conferences and journals, including TPAMI, TCSVT, TIP, TMM, TOMM, PR, CVPR, ICCV, ECCV, ICML, NeurIPS, AAAI, ACM MM, MMSP, PCS, ICME, and others.
Abstract
Learning-based visual data compression has recently achieved remarkable rate-distortion performance, surpassing that of traditional codecs. Looking ahead, I believe future research can incorporate broader structured data types together with image and video , leveraging the power of MLLMs and generative models to unlock new possibilities for next-generation video compression systems.
The first part of this talk establishes a foundational understanding of state-of-the-art learned compression for images and video in recent years. I will also discuss several emerging approaches, including diffusion model-based generative compression, knowledge distillation for lightweight design, and linear attention blocks as alternatives to transformers and convolutions, for enhanced rate-distortion-complexity performance.
The second part extends these principles to diverse modalities, covering multi-view representations (Gaussian Splatting, NeRF), sequential data (text, speech, tactile), and joint frameworks that share coding tools across data types to enable high coding efficiency and unified design.
Finally, the talk concludes with ongoing efforts toward unified coding frameworks that can support immersive communication and enable more comprehensive multimodal representation.
VSPC TC Rising Star Program
For more information about the VSPC Rising Star Program, please visit the official IEEE CAS VSPC Technical Committee page.
Goal
The Visual Signal Processing and Communications (VSPC) Rising Star Program was established in 2023 by the IEEE Circuits and Systems Society VSPC Technical Committee (TC) to annually recognize experts in emerging domains of visual signal processing and communications for their accomplishments and vision.
The Visual Signal Processing and Communications Rising Star Award will be presented annually at the IEEE Visual Communications and Image Processing (VCIP) conference. Selected candidates will be invited to deliver a VSPC Rising Star talk.
Through this talk, candidates will present their accomplishments and vision, and engage with VCIP attendees on the future of visual signal processing and communications research and applications. From the selected award candidates, one winner will ultimately receive the award. In exceptional circumstances, two winners may be selected at the same level.