| Image is an important carrier of the human cognition of external things. With the rapid development of computer and multimedia technology, the number of images increas-es explosively and it has become hot research problems in the area of pattern recognition to extract low-level features of the images for image classification and retrieval. However, compared to human visual system, computer vision is not in that state yet. Analyzing the low-level features which human brain concerns when it is doing image recognition using fMRI technologies and applying the results to computer vision are meaningful to promote cognitive abilities of computer vision.The main content of research in this thesis is to confirm the correspondences between neural activities and low-level features of the images by comparing the differences of neural activities in different regions when the subjects are watching different types of images. The detailed contents of the research are as follows:(1)Detect the functional activation areas processing for objects’shape, texture and color features. Take a set of images in different shape, texture and color as experiment stimuli and design a fMRI experiment. Collect the experiment data and do preprocessing to the data. Detect the functional activation areas processing for shape, texture and color features through statistical methods.(2)Identify the images’low-level features which the subjects pay attention to when they are watching different types of images. Get the correspondences between the neural activity and the low-level feature of the images when the human brain cognizes different types of images by computing the changes of neural activity within the region of inter-est when subjects viewing the face images, bus images, dinosaur images and mountain glaciers images.(3)When watching different types of images, subjects’neural activities within the region of interest are different. Using the differences, classify and predict the human brain’s image cognitive state. |
PDF Full Text Request