• Journal of IKEEE
• /
• v.28 no.2
• /
• pp.209-220
• /
• 2024

A sketch is an intuitive means to express information, but compared to actual images, it has the problem of being highly abstract, diverse, and sparse. Recent advances in deep learning models have made it possible to discover features that are common to images and sketches. In this paper, we summarize recent trends in sketch-based image retrieval (SBIR) but it is not limited to SBIR. Besides SBIR, we also introduce sketch-based image recognition and generation studies. Zero-shot learning enables models to recognize categories not encountered during training. Zero-shot SBIR methods are also discussed. Commonly used free-hand sketch datasets are summarized and retrieval performance based on these datasets is reported.

• Journal of KIISE:Databases
• /
• v.36 no.4
• /
• pp.263-279
• /
• 2009

Shape features in a content-based image retrieval (CBIR) system are divided into two classes: contour-based and region-based. Contour-based shape features are simple but they are not as efficient as region-based shape features. Most systems using the region-based shape feature have to extract the region firs t. The prior works on region-based systems still have shortcomings. They are complex to implement, particularly with respect to region extraction, and do not sufficiently use the spatial relationship between regions in the distance model In this paper, a region extraction method that is the combination of an edge-based method and a region growing method is proposed to accurately extract regions inside an object. Edges inside an object are accurately detected based on the Canny edge detector and the Hough transform. And the modified Integrated Region Matching (IRM) scheme which includes the adjacency relationship of regions is also proposed. It is used to compute the distance between images for the similarity search using shape features. The experimental results show the effectiveness of our region extraction method as well as the modified IRM. In comparison with other works, it is shown that the new region extraction method outperforms others.

• Journal of KIISE:Databases
• /
• v.33 no.4
• /
• pp.372-383
• /
• 2006

Among the approaches to image database management, content-based image retrieval (CBIR) is viewed as having the best support for effective searching and browsing of large digital image libraries. Typical CBIR systems allow a user to provide a query image, from which low-level features are extracted and used to find 'similar' images in a database. However, there exists the semantic gap between human visual perception and low-level representations. An effective methodology for overcoming this semantic gap involves relevance feedback to perform feature re-weighting. Current approaches to feature re-weighting require the number of components for a feature representation to be the same for every image in consideration. Following this assumption, they map each component to an axis in the n-dimensional space, which we call the metric space; likewise the feature representation is stored in a fixed-length vector. However, with the emergence of features that do not have a fixed number of components in their representation, existing feature re-weighting approaches are invalidated. In this paper we propose a feature re-weighting technique that supports features regardless of whether or not they can be mapped into a metric space. Our approach analyses the feature distances calculated between the query image and the images in the database. Two-sided confidence intervals are used with the distances to obtain the information for feature re-weighting. There is no restriction on how the distances are calculated for each feature. This provides freedom for how feature representations are structured, i.e. there is no requirement for features to be represented in fixed-length vectors or metric space. Our experimental results show the effectiveness of our approach and in a comparison with other work, we can see how it outperforms previous work.

• Journal of Broadcast Engineering
• /
• v.7 no.2
• /
• pp.145-152
• /
• 2002

As the number of digital images available to users is exponentially growing due to the rapid development of digital technology, content-based image retrieval (CBIR) has been one of the most active research areas. A variety of image retrieval methods have been proposed, where, given an input query image, the images that are similar to the input are retrieved from an image database based on low-level features such as colors and textures. However, most of the existing retrieval methods did not consider the case when an input query image is a part of a whole image in the database due to the high complexity involved in partial matching. In this paper, we present an efficient method for partial image matching by using the histogram distribution relationships between query image and whole image. The proposed approach consists of two steps: the first step prunes the search space and the second step performs block-based retrieval using partial image matching to rank images in candidate set. The experimental results demonstrate the feasibility of the proposed algorithm after assuming that the response tune of the system is very high while retrieving only by using partial image matching without Pruning the search space.

• Journal of Korea Multimedia Society
• /
• v.7 no.9
• /
• pp.1312-1320
• /
• 2004

Recently, even if a lot of visual feature representations have been studied and systems have been built, there is a limit to existing content-based image retrieval mechanism in its availability. One of the limits is the gap between a user's high-level concepts and a system's low-level features. And human beings' subjectivity in perceiving similarity is excluded. Therefore, correct visual information delivery and a method that can retrieve the data efficiently are required. Relevance feedback can increase the efficiency of image retrieval because it responds of a user's information needs in multimedia retrieval. This paper proposes an efficient CBIR introducing positive and negative relevance feedback with threshold value and pre-fetching to improve the performance of conventional relevance feedback mechanisms. With this Proposed feedback strategy, we implement an image retrieval system that improves the conventional retrieval system.

• Journal of Korea Multimedia Society
• /
• v.10 no.11
• /
• pp.1398-1406
• /
• 2007

Providing natural and efficient access to the fast growing multimedia information is a critical aspect for content-based information system. Query by humming system allows the user to find a song by humming part of the tune form music database. Conventional music information retrieval systems use a high precision pitch extraction method. However, it is very difficult to extract true pitch perfectly. So, In this paper, we propose to use YIN parameter with applying the reliability to reduce the pitch extraction errors. And we describes developed music information retrieval method based on a query by humming system which uses reliable feature extraction. Developed system is based on a continuous dynamic programming algorithm with features including pitch, duration and energy along with their confidence measures. The experiment showed that the proposed method could reduce the errors of the top-10 7.2% and the top-1 9.1% compared with the cepsturm based multiple pitch candidate. The overall retrieval system achieved 92.8% correct retrieval in the top-10 rank list on a database of 155 songs.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.43 no.1 s.307
• /
• pp.27-34
• /
• 2006

In this paper, we propose a combined region-based image retrieval system using EHD(Edge Histogram Descriptor) and CLD(Color Layout Descriptor) of MPEG-7 descriptors. The combined descriptor can efficiently describe edge and color features in terms of sub-image regions. That is, the basic unit for the selection of the region-of-interest (ROI) in the image is the sub-image block of the EHD, which corresponds to 16 (i.e., $4{\times}4)$ non-overlapping image blocks in the image space. This implies that, to have a one-to-one region correspondence between ELE and CLD, we need to take an $8{\times}8$ inverse DCT (IDCT) for the CLD. Experimental results show that the proposed retrieval scheme can be used for image retrieval with the ROI based image retrieval for MPEG-7 indexed images.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.40 no.2
• /
• pp.1-8
• /
• 2003

This paper proposes an object-based image retrieval technique based on the dominant color pair information. Most of existing methods for content based retrieval extract the features from an image as a whole, instead of an object of interest. As a result, the retrieval performance tends to degrade due to the background colors. This paper proposes an object based retrieval scheme, in which an object of interest is used as a query and the similarity is measured on candidate regions of DB images where the object may exist. From the segmented image, the dominant color pair information between adjacent regions is used for selecting candidate regions. The similarity between the query image and DB image is measured by using the color correlogram technique. The dominant color pair information is robust against translation, rotation, and scaling. Experimental results show that the performance of the proposed method has been improved by reducing the errors caused by background colors.

• Proceedings of the KAIS Fall Conference
• /
• 2003.11a
• /
• pp.223-230
• /
• 2003

Feature-based similarity retrieval has become an important research issue in multimedia database systems. The features of multimedia data are useful for discriminating between multimedia objects (e 'g', documents, images, video, music score, etc.). For example, images are represented by their color histograms, texture vectors, and shape descriptors, and are usually high-dimensional data. The performance of conventional multidimensional data structures(e'g', R- Tree family, K-D-B tree, grid file, TV-tree) tends to deteriorate as the number of dimensions of feature vectors increases. The R＊-tree is the most successful variant of the R-tree. In this paper, we propose a SOM-based R＊-tree as a new indexing method for high-dimensional feature vectors.The SOM-based R＊-tree combines SOM and R＊-tree to achieve search performance more scalable to high dimensionalities. Self-Organizing Maps (SOMs) provide mapping from high-dimensional feature vectors onto a two dimensional space. The mapping preserves the topology of the feature vectors. The map is called a topological of the feature map, and preserves the mutual relationship (similarity) in the feature spaces of input data, clustering mutually similar feature vectors in neighboring nodes. Each node of the topological feature map holds a codebook vector. A best-matching-image-list. (BMIL) holds similar images that are closest to each codebook vector. In a topological feature map, there are empty nodes in which no image is classified. When we build an R＊-tree, we use codebook vectors of topological feature map which eliminates the empty nodes that cause unnecessary disk access and degrade retrieval performance. We experimentally compare the retrieval time cost of a SOM-based R＊-tree with that of an SOM and an R＊-tree using color feature vectors extracted from 40, 000 images. The result show that the SOM-based R＊-tree outperforms both the SOM and R＊-tree due to the reduction of the number of nodes required to build R＊-tree and retrieval time cost.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.4
• /
• pp.273-280
• /
• 2013

This paper presents novel OCS-LBP (Oriented Center Symmetric Local Binary Patterns) based on orientation of pixel gradient and image retrieval system based on BoF (Bag-of-Feature) and random forest classifier. Feature vectors extracted from training data are clustered into code book and each feature is transformed new BoF feature using code book. BoF features are applied to random forest for training and random forest having N classes is constructed by combining several decision trees. For testing, the same OCS-LBP feature is extracted from a query image and BoF is applied to trained random forest classifier. In contrast to conventional retrieval system, query image selects similar K-nearest neighbor (K-NN) classes after random forest is performed. Then, Top K similar images are retrieved from database images that are only labeled K-NN classes. Compared with other retrieval algorithms, the proposed method shows both fast processing time and improved retrieval performance.