EDIS - Efficient Database Image Search

The goal of this thesis is to implement an efficient content based image database retrieval. The first stage of this project is to match a particular image to a database which might contain thousands of images. The advanced stage is to match an image region from a video frame to an image on a database, in real-time if possible.

The choice of image representation and image matching is key for the efficiency of this algorithm, the matching will not be done using the whole image data (pixels). At an initial stage, the image database might be local to the image retrieving device, in order not to account for data transmission latency. The outcome of the algorithm might be a small set of the most similar images.

The outcome of this project is key to scale an application which needs to query a database of images and fetch a small set of very similar images, if they are present.

The expected result is a component which receives an image representation extracted from a video frame and outputs a list of similar image representations present in a database. The component must do this task as efficient and as accurate as possible.

 

For any additional information regarding this project, please contact us using the inquiries form.