close

The Department of Computer Science at the University of Cyprus cordially invites you to the Colloquium entitled:

Rank based Estimation of Spectral Sensitivity Functions

 

Speaker: Dr. Maryam Darrodi
Affiliation: University of East Anglia and University College London, United Kingdom
Category: Colloquium
Location: Room 148, Faculty of Pure and Applied Sciences (FST-01), 1 University Avenue, 2109 Nicosia, Cyprus (directions)
Date: Wednesday, October 7, 2015
Time: 13:00-14:00 EET
Host: Constantinos Pattichis (pattichi-AT-cs.ucy.ac.cy)
URL: https://www.cs.ucy.ac.cy/colloquium/index.php#cs.ucy.2015.darrodi

Abstract:
The spectral sensitivity functions of a colour device e.g. a digital camera need to be known in order to accurately predict its response to spectral stimuli. These functions can be determined by cumbersome measurements in the lab or through statistical inference. The latter is the main focus of this paper. Typically, spectral sensitivities are estimated through linear regression assuming camera responds linearly to spectral stimuli. However, for rendered images (e.g. non-raw images taken by a mobile phone), this assumption is violated which can significantly impact an accuracy of sensor estimation. Here we describe a novel camera spectral sensitivity estimation technique that is robust for both linear and non-linear images. We note that the rank order of the sensor raw outputs should be the same as the rank order of the final rendered non-linear image assuming that the applied non-linear functions (gamma, camera curve etc.) are monotonically increasing. Each rank order can be written as a linear inequality that defines a linear constraint. These can be combined with other constraints such as zero sensitivity at ends of spectrum, unity of integration and linear combination of basis functions into the regression formulation. Solving for device sensitivities by this constrained regression is a quadratic programming problem. Our results show that the proposed rank-based method provides estimated spectral sensitivity functions that are consistently plausible for both raw and rendered data.

Short Bio:
Maryam Darrodi is a PostDoc of Colour Science at the University of East Anglia and University College London. She joined UEA in 2012 when she was awarded Senior Research Associate role at the age of 26. Maryam received her bachelors degree in Statistics first at the University of Beheshty (Tehran, Iran) and then directly proceeded to her doctoral degrees at University of Leeds in the UK where she was awarded PhD from school of Design. Maryam's current research is funded by Apple, Unilever, ESPRC, Buhler-Sortex and DataColor. Her overall interests is in different aspects of applied computer science, colour physics, psychophysics and psychology.

  Web: https://www.cs.ucy.ac.cy/colloquium/
  Mailing List: https://listserv.cs.ucy.ac.cy/mailman/listinfo/cs-colloquium
  RSS: https://www.cs.ucy.ac.cy/colloquium/rss.xml
  Calendar: https://www.cs.ucy.ac.cy/colloquium/schedule/cs.ucy.2015.darrodi.ics

Sponsor: The CS Colloquium Series is supported by a generous donation from Microsoft