A generic scheme for color image retrieval based on the multivariate Wald-Wolfowitz test

Abstract

In this study, a conceptually simple, yet flexible and extendable strategy to contrast two different color images is introduced. The proposed approach is based on the multivariate Wald-Wolfowitz test, a nonparametric test that assesses the commonality between two different sets of multivariate observations. It provides an aggregate gauge of the match between color images, taking into consideration all the (selected) low-level characteristics, while alleviating correspondence issues. We show that a powerful measure of similarity between two color images can emerge from the statistical comparison of their representations in a properly formed feature space. For the sake of simplicity, the RGB-space is selected as the feature space, while we are experimenting with different ways to represent the images within this space. By altering the feature-extraction implementation, complementary ways to portray the image content appear. The reported results, from the application on a diverse collection of images, clearly demonstrate the effectiveness of our method, its superiority over previous methods, and suggest that even further improvements can be achieved along the same line of research. It is not only the unifying character that makes our strategy appealing, but also the fact that the retrieval performance does not increase continuously with the amount of details in the image representation. The latter sets an upper limit to the computational demands and reminds of performance plateaus reached by novel approaches in information retrieval.

Texture retrieval based on a non-parametric measure for multivariate distributions

“…Not only different image characteristics can -in principle -be combined naturally in one type of query, but also different types of queries can evolve independently and their results can be compared across types. The latter is a direct consequence of the fact that the measured W-index relates directly to significance-level and therefore can be used as an absolute measure to rank among the results of different types of query [13].…”

“…Given the feature extraction step (analysed in the following Section), a representative set of textural characteristics -formed as feature vectors -is selected for a couple of texture images that are going to be compared. W is then computed and used as a similarity measure in a way that the more positive its value is, the more similar the two images are [13]. The W-quantity computed between pairs of images plays the role of a "distributional distance" acting on samples of image constituents, and therefore inherits interesting invariant characteristics such as rotation and translation invariance.…”

Color Image Retrieval Based on Non-Parametric Statistical Tests of Hypothesis

“…To achieve this, the test for equality of means, i.e. Mann-Whitney U (MWU) test [23] is employed. In general, the actual pixel values either in the query image or the target image may exeed 20, thus value of U approaches Gaussian distribution, and thus the null hypothesis can be tested by Z test.…”

“…Theoharatos et al [23] proposed a system, based on multivariate non-parametric test, namely Wald-Wolfowitz test (WW-test), and graph theoretic framework of minimal-spanning-tree (MST). In this work, first, the MST is constructed based on the sample identities of the points taken from the images.…”

Color Spatial Arrangement for Image Retrieval by Visual Similarity

“…To this end, a number of models were experimented with that permit the representation and comparison of images in terms of quantitative indexes of visual features [3], [4], [5]. In particular, different techniques were identified and experimented with to represent the content of single images according to low-level features, such as color [6], [7], [8], texture [9], [10], shape [11], [12], [13], and structure [14], [15]; Color Image Processing intermediate-level features of saliency [16], [17], [18] and spatial relationships [19], [20], [21], [22], [23]; or high-level traits modeling the semantics of image content [24], [25], [26]. In doing so, extracted features may either refer to the overall image (e.g., a color histogram), or to any subset of pixels constituting a spatial entity with some apparent visual cohesion in the user's perception.…”