The Studies About Feature Integration Method Based On The Correlation Theory

Since Hubel and Wiesel pioneering work in the visual aspects in the 1960s, visual information processing started to be studied by the way-depth. In visual information processing, the study of feature integration is a major research directions. Paper is related to the theory and research methods of features integration in visual information processing system.On the classical view of the reduction, through the convergence and input of processing channel, which eventually led to the response of a high degree of specificity of individual nerve cells, and cells to complete the integration from the distribution of information. Such a"grandmother cell"was considered to be in a higher level of inspections, as Barlow will be the end of the processing of information. However, because the huge amount handled by the brain, for a single neuron model, due to plasticity, cannot to solve problems of the expansion of information, and meet the fast changing characterization.Hebb considered that the visual images should be reflected in the interaction between neurons, and not the characterization by a single neuronal activity. But Hebb model assumptions in the information process, only a small part of community is activated in specific visual cortex region, and other communities is inhibited, and that is"overlapping barriers"of Hebb model.In order to solve these problems, Singer and Gray offered a correlation theory hypothesis on visual features integration, which considering, the characteristics of different coding brain cells are linked through its activities in the relevant time. That is, synchronous release is a natural way to achieve this. Paper based on the correlation theory, relating to the theory and research methods of features integration in visual information processing system, design and implement a features integration methods based on Visual C++.In order to understand the characteristics of human visual system and how visual information processing, inspire the information processing methods of computer vision, the first chapter research and analysis of the human visual system physiological basis, as well as cognitive of perception organization and neural mechanisms. Thus to make the theoretical foundation to study computer technologies on visual information processing, as well as biological basis knowledge used by the features integration.In the test of the feasibility of the correlation theory, relating to early visual integration computable model and experimental, research and design two psychological experiments. Based on the results, affirmed the feasibility of visual information integration used by correlation theory, and established features integration methods of synchronous release.In the study and compare traditional features integration modules, as well as inspired by synchronous oscillation model and features bundling issues, based on the correlation theory, design and implement a new synchronous integrated approach: a synchronization integrated method based on Visual C++.According to integration mechanism in features bundling, and the basis for the existence of illusion integration phenomena, from another perspective, investigated the process of the attention integrated individually, scattered features into unified, the overall perception. On this basis, establish the divide rules of basic shape features in synchronization integration method, according to the basic shape in the modular design tools VC. And from the features bundling phenomenon, initially establish include classification feature extraction in the modular design.With the study of synchronous oscillation theory, as well as three synchronous oscillation models analysis and comparison, identify the two subsystems in synchronous integration module: the integration of overall features and extraction of classification features.Ultimate successful development and demonstration of a VC-based synchronization features integrated system, which could complete shape, color, and location information automatically read and integration, and extract classification shape features in overall feature image, and demonstrate its classification features image relative to the overall feature image, and prepare for the high-level processing and analysis. And adding two psychological experiments demonstrate modules, provide the theoretical basis for the system.Paper designed the synchronization features integration method based on VC, simplify the integrated module, and extract the features relating to feature corresponding code, synchronization feature integration process be seen as a feature reader and synchronous mapping, and the analysis and perceptual formed into a high-level processing units. Thus reduce the complexity of the module, more convenient to get the feature image corresponding to the original image, and can demonstrate the classification features extracted from the overall feature image, to summary the similar features, preparing for the high-level processing, and the better realize the integrated thinking of correlation theory.In the development of design process, encountered two major problems, one is the addition of new feature, and the other is hierarchical display of similar features. Address these two issues, the main consideration solutions as follows. For the addition of new features, using method of manually expansion of feature code, when an external module read identifiers, processing mathematical nature and graphics rendering function design, establish their identity code into OnDraw function as a new feature added; and the other solution using multi-view in multi-document, making the different classification features image, appeared in a separate pop-up window view area, so that we can see a number of classification features at the same time, and further enhance the visual contrast.As a central issue of modern human visual perception study, images feature integrated theory and method will be research and development continue in-depth. Among the significance not only provides the feasibility of experimental basis for correlation theory, but more importantly, for a new try for synchronous integration technology tools and platform design, for the further research and application of future synchronization integration methods provide a viable and meaningful theory and practice basis.