Psychovisual Perception Scale Based on a Neural Network

EasyChair Preprint 3834, version 1

Abstract

The purpose of this article is to construct a psychophysical scale of visual perception from lighting scene based on a direct propagation neural network using for assessment of real or synthesized images with a space brightness distribution.

The difference between brightness in the lighting scene causes a sensation from "barely noticeable" to "painful" for the observer. In this work, the neural network acts as an" expert " who can evaluate these sensations. To train a neural network, input parameters were used: the brightness of the light source and background, and output parameters: the sensations of observers. These data were obtained at the experimental installation of the Department of lighting engineering of the MPEI (NRU) with the involvement of 10 observers. During the experiment, instructions were worked out for the unified definition of sensations and selected a model of visual perception scale with five categories.

The results obtained during the experiment are agreed with the numerical scale for evaluating of visual perception of lighting proposed by Lekish and Holladay. Based on these data, Neural network was train to predict a sensation at the level of 40-70%, depending on the scale category. The lowest probability of prediction was calculated for the categories "comfort" and "discomfort" that shows that the input data were insufficient. A new experiment should be done with correct calibration of the equipment at each level of brightness of adaptation and considering the instructions of observers for each sensation. This work has shown that idea to build a scale for evaluating the comfort of the spatial distribution of brightness in a lighting scene can be used. To do this experimental sample sufficient for training a neural network should be provided.

The novelty consists in using a neural network as an expert to assess the degree of comfort of the lighting scene.

Keyphrases: glare discomfort, lighting quality, neural network, scale of visual perception

@booklet{EasyChair:3834,
  author    = {Vladimir Budak and Ekaterina Ilyina},
  title     = {Psychovisual Perception Scale Based on a Neural Network},
  howpublished = {EasyChair Preprint 3834},
  year      = {EasyChair, 2020}}