Realization Law of Pragnanz and Closure of gestalt theory using Active Learning Method

Visual system in human is one of the most complicated and efficient kind of such systems. Visual sensing can produce the highest level of intelligent for live creatures and also machines[1]. Understanding, modelling, simulating and implementing of biological visual systems artificially is an important part of cognitive neuroscience, also one of the most important steps toward devising adaptive intelligent machines is to understand the visual system of the human and reconstruct a similar system artificially. Gestalt theory and related psychological theories try to explain how human’s brain can perform cognitive and computational tasks. Gestalt psychology tries to understand the laws of our ability to acquire and maintain meaningful perceptions in an apparently chaotic world.They claimed that humans perceive an overall picture from the environment as a whole patternin a way that individual parts have no meaning but by combining independent parts of the whole they can reveal an identifiable pattern.Gestalt theorists developed rules of perception to explain their ideas include Law of Pragnanz, Similarity, Closure and etc. Many researchers try to duplicate vision in computational systems. In this paper by using ALM along with morphology concept the law of Pragnanz and Closure have been developed. There are some methods to do image restoration like inpainting[2]. Convolutional Neural Networks (CNN) modeled to recognize shapes according to previous training. These networks can distinguish noisy shapes in many cases[3]. Anistropic Contour Completion (ACC) methods present edge grouping algorithms for finding a closed patterns starting from a particular edge point[4]. In this article a novel algorithmis introduced for completing imperfect or defective images based on patterns and symmetry of rest of shapes. In this method we can specify accuracy requirement. This algorithm extracts incomplete patterns and predicts removed parts after all clusters them in an unsupervised manner. Here we are concerned on how to extract and “fill-in”, once they have been selected.Regions are automatically filled with the structure of their surroundings, in a form that will be explained later in this paper.

Visual system in human is one of the most complicated and efficient kind of such systems. Visual sensing can produce the highest level of intelligent for live creatures and also machines[1]. Understanding, modelling, simulating and implementing of biological visual systems artificially is an important part of cognitive neuroscience, also one of the most important steps toward devising adaptive intelligent machines is to understand the visual system of the human and reconstruct a similar system artificially. Gestalt theory and related psychological theories try to explain how human’s brain can perform cognitive and computational tasks. Gestalt psychology tries to understand the laws of our ability to acquire and maintain meaningful perceptions in an apparently chaotic world.They claimed that humans perceive an overall picture from the environment as a whole patternin a way that individual parts have no meaning but by combining independent parts of the whole they can reveal an identifiable pattern.Gestalt theorists developed rules of perception to explain their ideas include Law of Pragnanz, Similarity, Closure and etc. Many researchers try to duplicate vision in computational systems. In this paper by using ALM along with morphology concept the law of Pragnanz and Closure have been developed. There are some methods to do image restoration like inpainting[2]. Convolutional Neural Networks (CNN) modeled to recognize shapes according to previous training. These networks can distinguish noisy shapes in many cases[3]. Anistropic Contour Completion (ACC) methods present edge grouping algorithms for finding a closed patterns starting from a particular edge point[4]. In this article a novel algorithmis introduced for completing imperfect or defective images based on patterns and symmetry of rest of shapes. In this method we can specify accuracy requirement. This algorithm extracts incomplete patterns and predicts removed parts after all clusters them in an unsupervised manner. Here we are concerned on how to extract and “fill-in”, once they have been selected.Regions are automatically filled with the structure of their surroundings, in a form that will be explained later in this paper.