This study attempts to provide fast indexing technique which will help to retrieve images from the database quickly and focuses on how to retrieve most relevant images from the database. The need for efficient Content-based Image Retrieval (CBIR) has increased tremendously in many application areas such as biomedicine, military, commerce, education and web image classification and searching. The semantic gap is the greatest challenge in the CBIR. The semantic gap is the lack of coincidence between the information that one can extract from the visual data and the interpretation that the same data have for a user in a given situation. The CBIR uses the visual contents of an image such as color, shape, texture and spatial layout to represent and index the image. In typical content-based image retrieval systems, the visual contents of the images in the database are extracted and described by multi-dimensional …

In this paper we presented an integrated schema for semantic object segmentation and content-based object search based on the region-based video object model. We first discussed AMOS, a generic video object segmentation system which combines low level automatic region segmentation with user input for defining and tracking semantic video objects. Our experiments and performance evaluation have shown very good segmentation results. Using the region-based video object model, an object query model which effectively combines local region-level features and spatial-temporal structures is then presented. Experiments have shown promising results and great potential for developing advanced video search tools for semantic video representations such as MPEG-4. In the future work, we will include multiple objects tracking, which can potentially be supported by our current algorithms. It would also be desirable to develop a hybrid strategy where video objects with simple motion and background can be tracked faster with less computation. Building upon the visual searching tools of video objects, we will explore object level video search and study application of the object search tools to MPEG-4 scene descriptions (i.e., BIFS). ate this problem and improve searching accuracy, we develop a novel query time region merging process (Figure 8). For each candidate region list, the query system will merge regions from a same video object into a large "virtual region" if 1) they spatially connected with each other; 2) the merged region is closer to the query region (feature distance). When a "virtual region" is generated, it is added to the current candidate region list. 3) Perform join (outer join) of the region lists on ObjectID to create a candidate object list. Each candidate object in turn contains a list of regions. A "NULL" region is used when :-a region list doesn't contains regions with the ObjectID of a being-joined object-a region appears (i.e. matched) more than once in a being-joined object 4) Compute the distance between the query object and each object in the candidate object list as follows: + (3) where is the ith query region. is the ith region in a candidate object. FD(.) is the feature distance between a region and its corresponding query region. If is NULL, maximum distance (i.e., 1) is assigned. (spatial), (topological) and (temporal) are structure features of the query object. , and are retrieved from database using indices on ObjectID, RegionID. When there is a NULL region (due to the above join process), the …