Windsurf: region-based image retrieval using wavelets

Abstract

Abstract

SHIATSU: tagging and retrieving videos without worries

“…According to a recent survey, 1 video traffic over the Internet amounts nowadays to over 40% of globally transmitted data, surpassing peer-topeer traffic for the first time. Besides the obvious technological advances needed to store and transmit such ever-increasing amount of data, the problem of correctly indexing and categorizing video data is particularly important, in order to allow the development of efficient and user-friendly browsing tools.…”

A Feature Level Fusion in Similarity Matching to Content-Based Image Retrieval

“…Then each region can be described by means of local features. The overall similarity between two images is calculated based on all the corresponding region-based local features [5,4]. In this section, we propose a clustering-based image segmentation technique and an image to image level similarity matching function based on clusters or regions generated from the segmentation.…”

“…To represent each region with local features, we consider information on weight and color-texture as in Ref. [4]. of each region R i is a 3-D vector and represented by the k-means cluster center, i.e., the average value for each of the three color channels in HSV space of all the image blocks in this region.…”

“…CBIR systems computes relevance of query and database images based on the visual similarity of low-level (e.g., color, texture, shape, edge etc.) image features [1,2,3,4,5,6]. Over the last few years, research interest into CBIR has grown rapidly from database, computer vision, pattern recognition, machine learning, and information retrieval communities.…”

“…To retain spatial information, some fixed grid based image partitioning techniques have been proposed with moderate success toward semantic-based retrieval [2]. Recently, region-based image retrieval (RBIR) have been studied widely [3,4,5,6], where each image is first segmented automatically into variable number of homogeneous regions, features of each regions are extracted and images are retrieved based on the similarity between regions. Although, there is a strong correlation between segmented regions and real-world objects, accurate segmentation of generalpurpose images is still an unsolved problem [5].…”