Traditionally, how to bridge the gap between low-level visual features and high-level semantic concepts has been a tough task for researchers. In this article, we propose a novel plausible model, namely cellular Bayesian networks (CBNs), to model the process of visual perception. The new model takes advantage of both the low-level visual features, such as colors, textures, and shapes, of target objects and the interrelationship between the known objects, and integrates them into a Bayesian framework, which possesses both firm theoretical foundation and wide practical applications. The novel model successfully overcomes some weakness of traditional Bayesian Network (BN), which prohibits BN being applied to large-scale cognitive problem. The experimental simulation also demonstrates that the CBNs model outperforms purely Bottom-up strategy 6% or more in the task of shape recognition. Finally, although the CBNs model is designed for visual perception, it has great potential to be applied to other areas as well.
TopIntroduction
Cognitive informatics (CI) is a transdisciplinary enquiry of cognitive and information sciences that investigates the internal information processing mechanisms and processes of the brain and natural intelligence (Wang, 2007). It covers a wide range of research fields, including the information-matter-energy (IME) model (Wang, 2003b), the layered reference model of the brain (LRMB) (Wang, Wang, Patel & Patel, 2006), the object-attribute-relation (OAR) model of information representation in the brain (Wang, 2006d; Wang & Wang, 2006), the cognitive informatics model of the brain (Wang, Liu, & Wang, 2003; Wang & Wang, 2006), natural intelligence (NI) (Wang, 2003b), autonomic computing (AC) (Wang, 2004), neural informatics (NeI) (Wang, 2002, 2003b, 2006a), CI laws of software (Wang, 2006b), the mechanisms of human perception processes (Wang, 2005a), the cognitive processes of formal inferences (Wang, 2005b), and the formal knowledge system (Wang, 2006c). Of all these branches, perception, as an interesting research field of CI, mainly focuses on how human beings perceive external world. Researchers have proposed an excellent model, the motivation/attitude-driven behavioral (MADB) model (Wang & Wang, 2006), to formally and quantitatively describe the relationship between the internal emotion, motivation, attitude, and the embodied external behaviors. In this article, we limit our work to visual perception, and propose a connectivity-based model to formally mimic the perceptual function of human beings.
The primary task of visual perceptual is to organize the visual features of an image into some already known objects. Yet, how to bridge the gap between low-level visual features and high-level semantic concept has long been a tough problem, which puzzles researchers all along. Until now, most of the proposed algorithms just focus on some particular objects, such as human faces, cars, people, and so forth (see e.g., Papageorgiou & Poggio, 2000; Schneiderman & Kanade, 2000; Tamminen a& Lampinen, 2003). Researchers utilize various schemes (see e.g., Broek et. al., 2005; Lai, Chang, Chang, Cheng, & Crandell, 2002; Tamminen & Lampinen, 2003]) to integrate the low-level visual features, including colors, textures and shapes into the profiles of target objects. Although some people (Murphy, Torralba, & Freeman, 2004) exploit background, or scene, information to improve the recognition, they do not take advantage of interrelationships between objects to help the identification process.