Image Classification and Querying Using Composite Region Templates

Bayesian fusion of camera metadata cues in semantic scene classification

“…Our baseline low-level classifier is similar to [13,16], extracting color histograms and wavelet (texture) features in a 4x4 block configuration, classifying each using an SVM, summing the outputs over all blocks, and shaping the sum using a sigmoid into a pseudo-probability. We trained it on an independent set of film and digital images not used elsewhere in this study.…”

“…Much research has been done on problems of scene classification [1,2,5,8,9,12,13,14,16,18]. The majority of these systems employed a learning-by-example approach based on low-level vision features derived exclusively from scene content.…”

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

“…In CBIR system, the researchers concentrate on developing low-level global visual features, namely color properties, shape, texture, and spatial relationship etc., which are used as query for the retrieval process [1][2][3][4]. The method proposed in [5-10; 10-15] classifies or segments the entire image into various regions according to the objects or structures present in the image, and the region-to-region comparison is made to measure the similarity between two images [3,11,12]. In a region-based system, the user has to provide one or more regions from the query image to start a query session.…”

Sunset scene classification using simulated image recomposition

“…This so-called "query by example" has often proved to be inadequate [3]. Knowing the category of a scene a priori helps narrow the search space dramatically, reducing the search time, increasing the hit rate, and lowering the falsealarm rate [4].…”

“…Belongie et al [1] reported 89% accuracy on a limited data set (approximately 60 training images and 30 testing images from each of twelve classes). Smith and Li [4] reported 86.1% accuracy on a small dataset of 91 training images (including 10 sunsets) and 266 testing images (including 36 sunsets). It is unclear how well any of the above systems would generalize for a large data set.…”