Development of image database systems by using spatial and temporal information

图像数据库系统不仅需要存储和管理大量的视觉信息，而且要为应用程序和用户提供良好的接口，以便有效地利用其数据。我们构建了亮度感知模型，该模型根据刺激/物理水平计算感知的亮度水平。然后，我们将在亮度感知中起关键作用的填充过程形式化并实现为局部迭代操作。仿真实验验证了对亮度虚幻现象的感知和对噪声阶跃边缘的检测，实验结果表明了该方法的有效性。在此基础上，构建了自然图像(水墨画)的目标提取和素描图像生成。第二年，提出了时空信息处理的计算视觉模型，并考虑了两者的融合。首先，我们开发了时间信息处理机制，并制定了运动元素积分器来进行运动角度的概率测量。然后，将复杂的视觉问题分解为以低维变量表示的局部过程，并对这些过程进行时空协同处理。最后，将本文提出的时空信息处理方法应用于大型遥感图像数据库系统，验证了该方法的有效性。

It is necessary for image database systems not only to store and manage many visual information but also to provide good interface adapted for applications and its users in order to utilize its data efficiently.At first year, we investigated spatial information processing method and applied our method for a specified image database (a database of Chinese brush paintings in ink). We constructed brightness perception model that computes perceived level of brightness from stimulus/physical level. Then, we formurated and implemented the filling-in process, which plays a key role in brightness perception, as locally iterative operation. Simulation experiments demonstrated perception of brightness illusory phenomena and detection of noisy step edges and its results showed the effectiveness of this method. At last, by using our method, the object extraction and sketch image generation from natural images (Chinese brush paintings in ink) was constructed.Second year, We proposed computational visual model for spatial and temporal information processing and considered the integuration of both of them. First, we developed temporal information processing mechanism and formulated motion element integrator that carries out probable measurement of motion angle. Then, complicated visual problems were decomposited into local processes represented in terms of low-dimensional variables, which were processed by cooperating spatial and temporal processing. At last, we applied our spatio-temporal information processing for large remote sensing image database systems and confirmed effectiveness of our proposed method.