Combining self-organizing neural nets with multivariate statistics for efficient color image retrieval

An efficient novel strategy for color-based image retrieval is introduced. It is a hybrid approach combining a data compression scheme based on self-organizing neural networks with a nonparametric statistical test for comparing vectorial distributions. First, the color content in each image is summarized by representative RGB-vectors extracted using the Neural-Gas network. The similarity between two images is then assessed as commonality between the corresponding representative color distributions and quantified using the multivariate Wald–Wolfowitz test. Experimental results drawn from the application to a diverse collection of color images show a significantly improved performance (approximately 10–15% higher) relative to both the popular, simplistic approach of color histogram and the sophisticated, computationally demanding technique of Earth Mover’s Distance.