550

Jasna Martinovic's Lab

Part of the Perception and Attention research theme at the School of Psychology, University of Aberdeen

  

Colour and Luminance in Object Representation

Early advantage of luminance for object representation and its cross-talk with chromatic channels in human visual scene analysis, a BBSRC funded project conducted by Jasna Martinovic (Principal Investigator) and Ben J Jennings (Post-doctoral researcher)

Detection and identification of objects are the most crucial goals of human visual perception. But the ease with which humans perceive objects in the environment is deceiving. The computations of visual signals in various brain regions which underlie object vision are anything but simple.

For example, during perception, different visual properties (e.g. luminance and colour) are initially sampled and processed by separate channels and undergo several stages of transformation in the parts of the brain that process visual signals. These separate channels in the visual system processes incoming information from the eyes in parallel, through a series of rapid hierarchically organized stages. This involves a cascade of neural activity that encompasses a whole range of brain areas, from sensory regions that represent simple features such as orientation to more broadly tuned brain regions that process contours, shapes and surfaces, ending with the regions that store our knowledge of the world.

Through these processes, a coherent representation of our environment segmented into objects that are layered in visual scenes is formed very rapidly - in less than 300 milliseconds of processing time - within a range of distributed brain areas.

In this project, it was our aim to investigate in which way the brain manages to coordinate the processing of information from different channels, such as colour and luminance, along each of the transformative stages that lead to the perception of objects. Both colour and luminance channels are crucial for everyday object vision but their contributions differ, with luminance considered to be more relevant for rapid processing of lines, edges, shape and motion and colour being more relevant for segmentation of visual scenes. But the extent to which colour and luminance channels function independently or interactively at different stages of visual processing in humans remains unknown even after many years of study, due to the difficulties in

1) Producing adequate stimuli for comparing low-, mid- and high-level visual processing along different channels

and

2) Analysing the differences in disparate, rapidly evolving neural processes that subserve normal human vision in the initial 300 miliseconds of processing time.

Our project attempted to overcome these problems through an innovative experimental approach which joined recordings of brain activity with a millisecond resolution together with the tools of colour psychophysics which can separate out different visual processing channels by precisely defining the chromatic and luminance properties of the stimuli.  We have created a large stimulus set of objects and matched 'non-objects' along the lines suggested by Sassi et al. (2010, iPerception, 1, 121-142), using spatial parameters of Gabors adjusted to deliver approximate performance along colour and luminance channels (see Wuerger and Morgan, 1999, JOSA A, 16, 436-444), and conducted a series of psychophysical experiments using these stimuli. For comparison, we also tested participants with stimuli designed to tap into the good continuation principle implemented in mid-level vision. Finally, we created a set of random Gabor patches comparable to these high and mid-level stimuli. Here are the links to the stimuli: line-drawings, Gaborised objects and non-objects, good continuation, poor continuation, random patches.


The time-course of cortical activations and their underlying generators was assessed with electroencephalography (EEG). Our initial study (Martinovic, Mordal and Wuerger, 2011, Journal of Vision, 11(7),1)  looked at a subset of signal combinations, but we have since tested all directions in DKL colour space psychophysically (Jennings, B.J. & Martinovic, J., 2014, Journal of Vision, 14(2), 9) and also looked at neural markers of some further combinations of chromatic and luminance signals using EEG. Finally, we performed some psychophysical experiments comparing  processing of objects with or without backgrounds  when driven by information from different channels and channel combinations. The manuscripts reporting these experiments are in the final stages of preparation.

We plan to continue research in this area, as findings of this study can provide an important insight into the ways in which the human brain utilises different types of information during visual processing and will thus significantly contribute to current knowledge on the relations between parallel processing of different features such as colour or luminance along the different hierarchical stages, from identification of differently oriented lines, to identification of familiar objects. Moreover, further study can add to the knowledge of neural mechanisms that allow preferential inputs of luminance information into object representation processes and thus enable it to drive everyday vision. In particular, it is important to understand how chromatic signals modulate neural responses driven by luminance and why this modulation might be useful for achieving optimal visual performance.

We're running several experiments related to this project as part of undergraduate projects. Summer placements are available in the lab for those students interested in visual perception and electroencephalography who would like to gain further research experience.

 
 
labImage






Contact the Martinovic Lab by email

School of Psychology
William Guild Building
University of Aberdeen
Aberdeen AB24 3UB, UK

Tel: +44 (0)1224 272240 (Dr Martinovic)
Fax: +44 (0)1224 273426