ByoungChul Ko

Integrated region-based image retrieval using region's spatial relationships

Among representative content-based image retrieval schemes, region-based retrieval has shown promise in retrieving similar images that exhibit considerable local variations. However, since humans are accustomed to relying on object-level concepts rather than low-level regions, robust and accurate object segmentation is an essential step. We propose a new multiple-region level image retrieval algorithm based on region-level image segmentation and its spatial relationship. To capture spatial similarity, we apply Hausdorff distance (HD) to our region-based image retrieval system, FRIP (finding region in the pictures). In contrast to other object or multiple region-based retrieval systems, we update classical HD to retrieve similar regions regardless of their spatial translation, insertion, and deletion. Furthermore, we incorporate relevance feedback to reflect the users high-level query and subjectivity to the system and to compensate for performance degradation due to imperfect image segmentation. The efficacy of our method is validated using a set of 3000 images from Corel-photo CD.

SVM-based salient region(s) extraction method for image retrieval

In region-based image retrieval, not all the regions are important for retrieving similar images and rather, the user is often interested in performing a query on only salient regions. Therefore, we propose a new method for extraction of salient regions using support vector machines (SVM) and a method for importance score learning according to the users interaction. Once an image is segmented, our algorithm permits the attention window (AW) according to the variation of an image and selects salient regions by using the pre-defined feature vector and SVM within the AW. By using SVM, we do not need to determine the heuristic feature parameters and produce more reasonable results. The distance values from SVM are used for initial importance scores of salient regions and our proposed updating algorithm using relevance feedback updates them automatically. Through performance comparison with parametric salient extraction method, our proposed method shows better performance as well as semantic query interface for object-level image retrieval.

Region-based Image Retrieval using probabilistic feature relevance learning

Abstract:Region-Based Image Retrieval (RBIR), a specialisation of content-based image retrieval, is a promising and important research area. RBIR usually requires good segmentation, which is often difficult to achieve in practice for several reasons, such as varying environmental conditions and occlusion. It is, therefore, imperative to develop effective mechanisms for interactive, region-based visual query in order to provide confident retrieval performance. In this paper, we present a novel RBIR system, Finding Region In the Pictures (FRIP), that uses human-centric relevance feedback to create similarity metric on-the-fly in order to overcome some of the limitations associated with RBIR systems. We use features such as colour, texture, normalised area, shape and location, extracted from each region of a segmented image, to represent image content. For each given query, we estimate local feature relevance using probabilistic relevance model, from which to create a flexible metric that is highly adaptive to query location. As a result, local data densities can be sufficiently exploited, whereby rapid performance improvement can be achieved. The efficacy of our method is validated and compared against other competing techniques using real world image data.

Region-based image retrieval system using efficient feature description

In this paper we introduce a region-based image retrieval system, FRIP. This system includes a robust image segmentation scheme using scaled and shifted color and shape description scheme using modified radius-based signature. For image segmentation, by using our proposed circular filter, we can keep the boundary of object naturally and merge small senseless regions of object into a whole body. For efficient shape description, we extract 5 features from each region: color, texture, scale, location, and shape. From these features, we calculate the similarity distance between the query and database regions and it returns the top K-nearest neighbor regions.

Image retrieval using flexible image subblocks

In this research, we propose a flexible subblock image retrieval algorithm which is robust to object translation, lighting change, object appearance or disappearance in an image which is divided into 9 non-overlapping subblocks. Furthermore, using Ohta-color moments and biorthogonal wavelet frames from each subblock, we can reduce the dimension of color space and improve the performance. Also, as the two features are applied to the multistep k-nearest neighbor algorithm, this system clearly outperforms the global color histogram, R G B moments, and HS1 moments using the same method. In addition, we provide another retrieval environment for the user which is a relative block location search as well as an absolute block location search. In the case of the block location search, the user selects a block the user wants to search.

Probabilistic neural networks supporting multi-class relevance feedback in region-based image retrieval

There are several relevance feedback algorithms available, some algorithms use ad-hoc heuristics or assume that feature vectors are independent regardless of their correlation. In this paper, we propose a new relevance feedback algorithm using probabilistic neural networks (PNN) supporting multi-class learning. In our approach, there is no need to assume that feature vectors are independent and it permits system to insert additional classes for detail classification. In addition, it does not take long computation time for training, because it has only four layers. In PNNs classification process, we keep the users entire past feedback actions as history in order to improve the performance for future iterations. In the history, our approach can capture the users subjective intension more precisely and prevent retrieval performance from fluctuating or degrading in the next iteration. To validate the effectiveness of our feedback approach, we incorporate the proposed algorithm to our region-based image retrieval tool FRIP (finding region in the pictures). The efficacy of our method is validated using a set of 3000 images from Corel-photo CD.

Multiple Regions and Their Spatial Relationship-Based Image Retrieval

In this paper, we present a new multiple regions and their spatial relationship-based image retrieval method. In this method, a semantic object is integrated as a set of related regions based on their spatial relationships and visual features. In contrast to other ROI (Region-of-Interest) or multiple region-based algorithms, we use the Hausdorff Distance (HD) to estimate spatial relationships between regions. By our proposed HD, we can simplify matching process between complex spatial relationships and admit spatial variations of regions, such as translation, rotation, insertion, and deletion. Furthermore, to solve the weight adjust problem automatically and to reflect users perceptual subjectivity to the system, we incorporate relevance feedback mechanism into our similarity measure process.

Query-by-Gesture: An Alternative Content-Based Image Retrieval Query Scheme

Abstract Current research on content-based image retrieval (CBIR) is centered on designing efficient query schemes in order to provide a user with effective mechanisms for image database search. Among representative CBIR query schemes, query-by-sketch has been one of the attractive query tools that are highly adaptive to users subjectivity. However, query-by-sketch has a few limitations. That is, most sketch tools demand expertise in image processing or computer vision of the user to provide good enough sketches that can be used as query. Furthermore, sketching the exact shape of an object using a mouse can be a burden on the user. To overcome some of the limitations associated with query-by-sketch, we propose a new query method for CBIR, query-by-gesture, that does not require sketches, thereby minimizing user interaction. In our system, the user does not need to use a mouse to make a sketch. Instead, the user draws the shape of the object that heshe intends to search in front of a camera by hand. In addition, our query-by-gesture technique uses relevance feedback to interactively improve retrieval performance and allow progressive refinement of query results according to the users specification. The efficacy of our proposed method is validated using images from the Corel-Photo CD.

Automatic text extraction in news images using morphology

In this paper we present a new method to extract both superimposed and embedded graphical texts in a freeze-frame of news video. The algorithm is summarized in the following three steps. For the first step, we convert a color image into a gray-level image and apply contrast stretching to enhance the contrast of the input image. Then, a modified local adaptive thresholding is applied to the contrast-stretched image. The second step is divided into three processes: eliminating text-like components by applying erosion, dilation, and (OpenClose + CloseOpen)/2 morphological operations, maintaining text components using (OpenClose + CloseOpen)/2 operation with a new Geo-correction method, and subtracting two result images for eliminating false-positive components further. In the third filtering step, the characteristics of each component such as the ratio of the number of pixels in each candidate component to the number of its boundary pixels and the ratio of the minor to the major axis of each bounding box are used. Acceptable results have been obtained using the proposed method on 300 news images with a recognition rate of 93.6%. Also, our method indicates a good performance on all the various kinds of images by adjusting the size of the structuring element.