Ontology Of Low Level Visual Objects
Blazej Skrzypulec
Matthews Suite, Maclaurin Building
About the event
Event type: Seminar
It is widely recognised in the cognitive sciences concerning vision that stimuli which affect the perceptual system are quite different from what we consciously see. The visual system is stimulated by light waves, but our visual field is usually filled with discrete visual objects. In the presentation, I consider the way in which basic, low-level visual objects are formed, and especially how they are constructed by transforming the structure of more primitive, non-object visual representations.
Various conceptions of objects’ structure have been proposed by the analytic theories. However, these philosophical conceptions have not been usually applied to visual objects. Even in philosophical works which investigate the process of constructing visual representations, the specific nature of structural change between non-object representations and basic visual objects is not usually explicitly analysed.
In the presentation, I start by proposing a processual criterion that allows to distinguish visual objects from other types of visual representations. In addition, I characterise the construction of low-level visual objects as the process in which the structure of non-object visual representation is extended.
Secondly, I describe the theories of early vision proposed by Clark and Raftopoulos, in order to investigate what kind of structural transition between non-object visual representation and low-level visual objects is assumed within them. According to Clark, visual objects are constructed when visual features are bound with visual locations by mechanisms of early vision. Within the Raftopoulos’s conception, visual objects are formed by preattentive segmentation and grouping processes.
Subsequently, I develop my conception of visual objects’ construction. I argue, that non-object visual representations are transformed into low-level visual objects when they stand in proper relational networks to other visual representations. I show that such relational notion is suggested by the scientific models of figure/ground discrimination in vision.
Both behavioural and neural models rely on the notion of competition between visual regions. From that a conclusion can be drawn, that the most important factors that allow to distinguish simple visual objects (figures) from non-object visual representations (that constitute ground) are not intrinsic features of visual regions but relations between them. Such a relational notion seems to be significantly different from approaches that can be found in conceptions by Clark and Raftopoulos.