Spatial and Temporal Image Database Navigation

The ubiquitous use of mobile devices nowadays causes large amounts of photos and videos. What is needed are tools that enable the user to gain an intuitive overview of the image quantity.

If large amounts of images are produced, the grouping of images, due to their visual content, is a very important sorting tool. The extraction of visual feature points and the calculation of corresponding descriptors form the basis of this analysis. The number of matching features of two images is a measure of their similarity.

In general, two similar pictures show the same scene from different camera perspectives. From the matching features of two similar images, the change in perspective is calculated in the form of a projective function. In this way, groups of similar images are arranged in a similarity graph. Images that are connected in the graph by an edge can be continuously merged into each other.

A specially developed navigation viewer allows spatial navigation within this similarity graph. From an initial image, triangular symbols are used to indicate all the images connected by a projective transformation. By moving the mouse over a symbol, the associated image is placed over the original image. With a mouse click, the connected image becomes the new original image. In this way you can navigate through the similarity graph.
 

 

An event viewer developed by us enables a quick overview of the temporal sequence of events. For this purpose, all images of one and the same motif are brought into a temporal sequence, through which you can navigate through time. Both viewers are web-based and can be integrated into any web interface.