Luca's Studio

I am a researcher in Computer Graphics & Human Perception

In this website I occasionally publish some info about my research and extra stuff that I do or I like, e.g. gaming prototypes, music, teaching, sport...

This "studio" is intended as a half way between a personal page and a blog. Other than content on computer science, it is going to contain flashes of various topics.

Research

My research focuses on creating models of different aspects of human vision to improve computer graphics techniques in terms of both quality and efficiency, mainly but not exclusively for AR / VR displays.
I have worked on peripheral vision, color, luminance-contrast, and depth. These models have been applied respectively to the following corresponding applications: foveated rendering (image and geometry), color grading, power saving, and 3D reconstruction.
I did my PhD in the Perception, Display and Fabrication group led by Prof. Piotr Didyk, at USI-IDSIA in Lugano, Switzerland. My doctoral thesis is available here. During my PhD, I had the opportunity to intern at Disney Research|Studios, where I worked on color perception, supervised by Dr. Tunç Aydin; and at Google, working on 3D reconstruction for stereoscopic rendering within the XR Core Tech Team, supervised by Prof. Cengiz Öztireli. Before that, I worked on the Air-Cobot project and obtained a MSc from University of Calabria (Italy) with a thesis focusing on emotion recognition (University of Plymouth, UK).

News

Publications

Temporal Brightness Management for Immersive Content

Luca Surace, Jorge Condor, Piotr Didyk
Eurographics Symposium on Rendering, 2025
Leveraging and extending low-level vision models to optimize energy consumption in VR displays, in two flavors: an optimization process minimizing contrast loss for offline content, and a PID control-inspired scheme for real-time rendering.
Project Page   Publisher Page

Gaze-Contingent Perceptual Level of Detail Prediction

Luca Surace, Cara Tursun, Ufuk Çelikcan, Piotr Didyk
Eurographics Symposium on Rendering, 2023
A technique to select the suitable LoD of a given mesh in the visual field. It keeps the same perceived quality and improves computational efficiency of a gaze-contingent system by saving on the rendered number of polygons. Tested also in combination with Nanite from Unreal engine.
Project Page   Publisher Page

Emotion Recognition in the wild using deep neural networks and Bayesian classifiers

Luca Surace, Massimiliano Patacchiola, Elena Battini Sonmez, William Spataro, Angelo Cangelosi
19th ACM International Conference on Multimodal Interaction (ICMI), 2017
A technique to recognize the emotion of a group of people from a picture taken in unconstrained settings. Based on standard CNN and Bayes' theorem, it infers an emotion category: positive, negative or neutral.
Code   Publisher Page

S-rep model for fundus image analysis

Francesco Calimeri, Claudio Stamile, Luca Surace
IEEE International Symposium on Signal Processing and Information Technology (ISSPIT), 2017
A 3D reconstruction of blood vessels from fundus eye images. The technique detects longitudinal changes in vessel structure using statistical analysis based on a generalized PCA approach.
  Publisher Page

Teaching

Bachelor Projects for USI Students

The objective of the bachelor's project is to develop an interactive tool for drawing 2D closed shapes. Each closed shape can be represented with a series of elliptic functions.
The higher the number of ellipses in the sequence, the more precise the approximation of the shape is (see the figure). The tool should, given the state of the current drawing:

  • compute its representation;
  • find similar shapes through comparison with several pre-computed representations of shapes;
  • suggest the similar shapes to the artist;
no machine learning is required.
The tool should also have basic commands for drawing, such as pencil, pen, eraser, undo and redo. A similar system is Autodraw. Possible extensions are clearly possible.
If interested, please contact me at my USI email.