Murat Birinci

Perceptual color descriptor based on spatial distribution: A top-down approach

Color features are the key-elements widely used in content-analysis and retrieval. However, most of them show severe limitations and drawbacks due to their inefficiency of modeling the human visual system with respect to color perception. Moreover, they cannot characterize all the properties of the color composition in a visual scenery. In this paper we present a perceptual color feature, which describes all major properties of prominent colors both in spatial and color domains. In accordance with the well-known Gestalt law, we adopt a global, top-down approach in order to model (see) the whole color composition before its parts and in this way we can avoid the problems of pixel-based approaches. In color domain the dominant colors are extracted along with their global properties and quad-tree decomposition partitions the image so as to characterize the spatial color distribution (SCD). We propose two efficient SCD descriptors; the proximity histograms, which distill the histogram of inter-color distances and the proximity grids, which cumulate the spatial co-occurrence of colors in a 2D grid. Both approaches are configurable and provide means of modeling SCD in a scalar and directional way. Combination of the extracted global and spatial properties forms the final descriptor, which is unbiased and robust to non-perceivable color elements in both spatial and color domains. Finally a penalty-trio model fuses all color properties in a similarity distance computation during retrieval. Experimental results approve the superiority of the proposed technique against powerful global and spatial color descriptors.

Neighborhood matching for object recognition algorithms based on local image features

Local image features around interest-points have been widely used in order to exploit the similarities between different views of an object in different images. While there are numerous algorithms on detecting the interest-points and defining the local features, few have focused on the importance of the matching process. In this paper, we presented a method that matches interest-points detected via any algorithm. The method is motivated from human perceptual rules, particularly the Gestalt Psychology, and realizes the fact that “The whole is different from the sum of its parts”. The efficacy of the algorithm is not only the ability to decrease the number of false positive matches but also to increase the number of true positives, yielding rock-steady results for any algorithm based on matching local features.

Neighborhood Matching for Image Retrieval

In the last few years, large-scale image retrieval has attracted a lot of attention from the multimedia community. Usual approaches addressing this task first generate an initial ranking of the reference images using fast approximations that do not take into consideration the spatial arrangement of local features in the image (e.g., the bag-of-words paradigm). The top positions of the rankings are then re-estimated with verification methods that deal with more complex information, such as the geometric layout of the image. This verification step allows pruning of many false positives at the expense of an increase in the computational complexity, which may prevent its application to large-scale retrieval problems. This paper describes a geometric method known as neighborhood matching (NM), which revisits the keypoint matching process by considering a neighborhood around each keypoint and improves the efficiency of a geometric verification step in the image search system. Multiple strategies are proposed and compared to incorporate NM into a large-scale image retrieval framework. A detailed analysis and comparison of these strategies and baseline methods have been investigated. The experiments show that the proposed method not only improves the computational efficiency, but also increases the retrieval performance and outperforms state-of-the-art methods in standard datasets, such as the Oxford 5 k and 105 k datasets, for which the spatial verification step has a significant impact on the system performance.

Perceptual color descriptor based on a spatial distribution model: Proximity histograms

Color is the main source of information particularly for content-analysis and retrieval. Most of the color descriptors, however, show severe limitations and drawbacks due to their incapability of modelling the human color perception. Moreover, they cannot characterize all the properties of the color composition in a visual scenery. In this paper we present a perceptual color feature, which describes all major properties of prominent colors both in spatial and color domain. In accordance with the well-known Gestalt law, we adopt a top-down approach in order to model (see) the whole color composition before its parts and in this way we can avoid the problems of pixel-based approaches. In color domain the dominant colors are extracted along with their global properties and quad-tree decomposition partitions the image so as to characterize the spatial color distribution (SCD). The proposed color model distils the histogram of inter-color distances. Combination of the extracted global and spatial properties forms the final descriptor, which is neither biased nor become noisy from the presence of such color elements that cannot be perceived in both spatial and color domains. Finally a penalty-trio model fuses all color properties in a similarity distance computation during retrieval. Experimental results approve the superiority of the proposed technique against well-known global and spatial descriptors.

Image color content description utilizing perceptual color correlogram

Most of the content-based image retrieval (CBIR) systems frequently utilize color as a discriminative feature among images, due to its robustness to noise, image degradations, and changes in resolution and orientation. While there are vast amount of color descriptors for describing global properties of colors such as ldquowhatrdquo and ldquohow muchrdquo color present in an image, less research has succeeded in describing the spatial properties among colors such as ldquowhererdquo and ldquohowrdquo. Color correlogram is one of the most promising spatial color descriptors at the current state of art. However, it has several limitations, which make it infeasible even for ordinary image databases. In this paper we present a perceptual approach that eliminates such restrictions from correlogram and further increases its discrimination power. Experimental results demonstrate correlogrampsilas handicaps and the success of the proposed approach in terms of retrieval accuracy and feasibility.

Dynamic queries with relevance feedback for content based image retrieval

A novel relevance feedback scheme utilizing dynamic queries for content based image retrieval systems is proposed, where the retrieval results are updated instantly based on the users feedback. The user is expected to label at least one image as positive or negative, revealing the gist of the expected retrieval results. Then the retrieval results are updated dynamically, without any further user interaction, based on the similarity of the query and the selected image in different feature spaces increasing the semantic accuracy of the retrieval. The proposed method not only invalidates the drawbacks of current relevance feedback systems in terms of user experience, but also provides an innovative stand point for the relevance feedback scheme as well.

A region-dependent image matching method for image and video annotation

In this paper we propose an image matching approach that selects the method of matching for each region in the image based on the region properties. This method can be used to find images similar to a query image from a database, which is useful for automatic image and video annotation. In this approach, each image is first divided into large homogeneous areas, identified as “texture areas”, and non-texture areas. Local descriptors are then used to match the keypoints in the non-texture areas, while texture regions are matched based on low level visual features. Experimental results prove that while exclusion of texture areas from local descriptor matching increases the efficiency of the whole process, utilization of appropriate measures for different regions can also increase the overall performance.

Directional Spatial Color Descriptor in a Perceptual Model: Proximity Grids

Most of the color features widely used in content-based image retrieval (CBIR) present severe limitations and drawbacks due to their inefficiency of modeling human visual system on color perception. Accordingly, they are not capable of characterizing both spatial and global properties of the color composition in visual scenery. In this paper, we present a perceptual color feature, which describes the global properties of the prominent colors along with a directional spatial descriptor, called as Proximity Grids. In color domain the dominant colors are extracted along with their global properties and quad-tree decomposition partitions the image so as to characterize the spatial color distribution (SCD). This approach is in accordance with the well-known Gestalt law, i.e. utilizing a top-down approach in order to model (see) the whole color composition before its parts and in this way we can avoid the problems of pixel-based approaches. The proximity grids, which cumulate the spatial co-occurrence of colors in a 2D grid, can successfully model the SCD of the prominent colors with respect to inter-color proximities and directions. Fusing both global and spatial properties forms the final descriptor, which is neither biased nor become noisy from the presence of such color elements, the so-called outliers that are not visible for humans in both spatial and color domains. Finally a penalty-trio model cumulates the differences among the color properties in a similarity distance computation during retrieval. Comparative evaluations against well-known global and spatial descriptors demonstrate the superiority of the proposed descriptor.