On Discrimination Power of Image Feature Vector

Abstract

AbstractImage Analysis using systems like Content based Image Retrieval is the information retrieval system which helps in retrieving similar images, found very useful in many areas. Feature vectors are the compact and precise representation of Images which effectively needs to be used in Image Retrieval System. The performance of Image Retrieval is majorly dependent on the quality of feature vectors. Many intelligent algorithms are proposed in literature to retrieve the features of images and represent it in an optimal way. Feature Vectors with reduced dimensions are always preferred for image retrieval systems. Quality of Feature Vectors is always checked using the output of image retrieval. There has always been a trade-off between the size of the feature vector and retrieval performance. Common measures used for checking retrieval system performance are precision, recall. This paper discusses the important measure: “discrimination power” of Image feature vector to check feature vector quality in context of image matching and retrieval. The discrimination power of Image Feature Vector, and its use-case with our feature extraction method is discussed here.KeywordsDiscrimination powerImage retrieval and analysisDimensionality reductionFeature extractionLow-level featuresImage analysis