Quantification of impacts of colour on affective quality of images

Most consumer images serve emotional functions as well as informational ones. The impression of an image can be affected not only by physical properties e. g. size, colour and media but also by the context of images, aesthetic properties and social/personal backgrounds of observers. However in traditional frameworks in image quality studies, the impacts of colour-appearance attributes on image quality have focused on maximising the informational functions of images, considering an image as a reproduced copy of a real scene. Thus, a new approach was adopted in this study in an attempt to investigate the emotional aspect of an image. The goal of this research is to study the impact of colour-appearance attributes of an image on emotional responses, and to develop quantitative models for predicting emotional response considering the context of the image. To achieve this goal, three sets of psychophysical and physiological experiments have been conducted. First, the relationship between colour-appearance attributes and overall affective response to images was investigated for four different types of image contents. It was found that image colourfulness and lightness contrast had a consistent influence on these relationships for all types of images. The relationships between emotional responses of image pleasantness and excitement were significantly different between positive images and negative images. Accordingly, quantitative models of image pleasantness and excitement were developed as a function of image colourfulness and contrast separately for the two groups of images. Finally, models of image pleasantness and excitement for positive and negative images were developed as a linear equation based on models developed for each colour attribute. The relationships between colour-appearance attributes and responses on colour-emotion scales, active-passive, heavy-light and warm-cool, were also studied for four different types of image content. Quantitative models of the three colour-emotion scales were developed as a function of colour attributes of images such as lightness, colourfulness and lightness contrast. As an application of using the colour-emotion model developed for images, the relationships between colouremotion scales and image emotion were investigated and quantitative models of image pleasantness and excitement were developed as functions of three colouremotion scales for two groups of images: positive and negative. The model performance based on the colour-emotion scales was compared with the performance of models based on the colour attributes. As a result, the latter model performed better than former. The impact of image content and colour attributes of an image on emotional responses to images was investigated by measuring physiological responses to images which were compared with the psychophysical responses. It was found that the activities in skin conductance and heart rate showed significantly greater responses for the images with personal meanings and significances. For the effect of colour attributes in images, it was found that more chromatic images generated higher activity in skin conductance responses. It was also found that lower contrast images generated higher activity in corrugator EMG responses