Research On Image Cognition-oriented Group Analysis Of FMRI Data

Image, a carrier of the complex information, is the main channel of computer access to information. Relative to the brain perception of the image, computer image cognition is at the primary stage. With the development of the fMRI technology, based on the prior knowledge in biology, this paper adopts group analysis method analyzing the fMRI data obtained in the brain cognitive processes. Then it establish effective links from the underlying image features to different kinds of images, as a result, getting an attention mechanism of brain to the image underlying characteristics. At last, applying the result to computer image classification and retrieval is expected to improve computer visual cognitive effects.The main content of this paper is contrasting, verifying and confirming perception regions of underlying characteristics’ shapes, textures and colors through the analysis and research of the group of fMRI data. Next it will get concern of brain to low-level features, when subjects watching different types of images, through the low-level features and the image recognition built by perceptual effective connection. The specific work is as follows:(1)Prepossess fMRI data, and one-sample t tests and conjunction comparative analysis method are adopted to locate active area of low-level features, such as shape, texture and color, in subjects’brains, as well as four types of images under the stimulus of activation of brain areas.(2)Use prior knowledge, and brain activation regions produced by individual function activation and t test group analysis are calculated, the ROI is redefined to remove the influences caused by individual brain differences in maximum, and extract the changing signals of ROI percentage, compare with the ROI that is defined by the activated coordinates which is get by using group analysis directly.(3)Variance analysis is used to analyze the interaction of two factors between the ROI and condition of stimulus (shapes, textures, colors and no change contrast), contrast, verify and confirm perception regions of shape, texture and color, then through four different types of image as a condition of stimulating and perceptual two-factor variance analysis, get the attention of image underlying characteristics, so as to implement the reverse research on the brain.