Everything looks better with lasers. This is a geometric calibration of our ultra-realistic 3D display (cl.cam.ac.uk/research/rainb…).

Our 2011 HDR_VDP-2 paper received the Test-of-Time Award from SIGGRAPH for "a significant and lasting impact on computer graphics and interactive techniques over at least a decade" blog.siggraph.org/2023/07/siggra… HDR-VDP-2 project page (updated): hdrvdp.sf.net

Merging multiple exposures into an HDR image often results in banding artefacts. We found that, in most cases, those artefacts are caused by inaccurate exposure times and that robust estimation of exposure times can eliminate those: doi.org/10.1109/TCI.20…

Neural denoising of path tracing can be improved with a trainable pyramid data structure. The trainable pyramid can extend the receptive field without introducing downsampling artifacts. Martin will be presenting a paper on that at #SIGGRAPH2023 - balint.io/nppd/

We have an open Research Associate (postdoc) position in AR/VR quality to work on perceptual representations, cost functions, and metrics for key visual computing applications in AR/VR. The closing date is 15 September 2023. jobs.cam.ac.uk/job/42566/

Any interesting ideas or results regarding Human Vision and Electronic Imaging? Consider submitting to #HVEI2024 (1000-word abstract or full paper) before 15 September. It is an excellent forum to present to like-minded experts. imaging.org/IST/Conference…

London Imaging Meeting this year is on Image Capture -- from photons to photos. The submission deadline is the 15th of March. imaging.org/IST/Conference…

New article in the Extended Reality Issue bit.ly/3PdJOLH by Maliha Ashraf Maliha Ashraf et al. Cambridge Computer Science build an analytical contrast sensitivity model, castleCSF, that explains consolidated contrast sensitivity data from 18 studies. doi.org/10.1167/jov.24…

I am happy to announce the release of ColorVideoVDP - a differentiable image and video quality that is built on the first principles of human color and spatiotemporal vision. Code: github.com/gfxdisp/ColorV… The work will be presented at #SIGGRAPH2024.

Congrats to Dounia Hammou for receiving the Best Student Paper Award at QoMEX 2025 for her work on the effect of viewing distance and display peak luminance on image quality and the new HDR-VDC dataset - cl.cam.ac.uk/research/rainb…

True 3D volumetric displays require the tracking of the eye position to align images presented in 3D space. At #ISMAR2024, we presented a technique that tracks accurate eye position and gaze direction and requires only one-time calibration. cl.cam.ac.uk/research/rainb…

Announcing our SIGGRAPH Asia ➡️ Hong Kong work with Rafal Mantiuk lab studying the perception of distortions in augmented reality (AR) - Our data will help researchers create accurate metrics for AR. Come see the talk on Dec 3rd, and check out the AR-DAVID project page: cl.cam.ac.uk/research/rainb…

Our #SIGGRAPHAsia2024 honourable mention award for our work on AR-DAVID. cl.cam.ac.uk/research/rainb…

London Imaging Meeting this year is on Lighting and Imaging. Paper submission deadline is 16 May 2025. iop.eventsair.com/lim-2025

I will be giving Optica Webinar on ColorVideoVDP: A Visual Difference Predictor for Image, Video and Display Distortions optica.org/events/webinar… via The Optical Society now Optica

Neural networks work better with HDR and RAW data represented in perceptually uniform color spaces. We showed that in our NITRE #CVPR2025 workshop paper "Training Neural Networks on RAW and HDR Images for Restoration Tasks" - arxiv.org/abs/2312.03640

Large vision models, such as DINOv2, start to see like us. We found that they have similar contrast discrimination and matching characteristics as human vision; i.e., their responses are consistent with psychophysical data. Details in our #CVPR2025 paper cl.cam.ac.uk/research/rainb…

ColorVideoVDP-ML was awarded second place in the full reference category of ICME Generalizable HDR & SDR Video Quality. The new metric models both the low-level (ColorVideoVDP) and high-level (ML) human vision. Now the code is available at github.com/gfxdisp/ColorV…

We have two open positions (pre-doc and post-doc levels) in my group in Cambridge to work on 3D video quality, in collaboration with Meta. Join us to shape the future of AR/VR media with impactful research. Post-doc: cam.ac.uk/jobs/research-… Pre-doc: cam.ac.uk/jobs/research-…

Is your ultra-HD TV worth it? Scientists measure the resolution limit of the human eye cam.ac.uk/research/news/…