Picture Dataset Research Articles

Fuzzy-linked phase congruency-based feature descriptors for image retrieval

ABSTRACTThe demand for acute and compact feature descriptors remains a key issue of concern for prevailing image retrieval schemes. This paper offers dual feature descriptors characterised by phase congruency (PC) and fuzzy logic for image indexing and retrieval. The proposed mechanism commences with colour space conversion of RGB query images to L*a*b* triplets and further application of PC generates relevant feature information. The ensuing visual features are blended by fuzzy rules to formulate the unified feature histograms and later fuzzy quantised to produce two feature descriptors termed as PC-based colour edge directivity descriptor (PC-CEDD), PC-based fuzzy colour texture histogram (PC-FCTH). The resulting descriptors occupy minimal storage space of 23–74 bytes per image, with 60% reduction in feature extraction time in comparison with CEDD, FCTH. Relative precision–recall and mean average precision (MAP) analysis of the intended feature histograms on medical, texture, and object picture dataset signify the improvement in retrieval performance. Furthermore, average normalised modified retrieval rank analysis of the intended descriptors reveals the better matching quality of the given query image.

The bank of standardized stimuli (BOSS): comparison between French and English norms

Throughout the last decades, numerous picture data sets have been developed, such as the Snodgrass and Vanderwart (1980) set, and have been normalized for variables such as name and familiarity; however, due to cultural and linguistic differences, norms can vary from one country to another. The effect due specifically to culture has already been demonstrated by comparing samples from different countries where the same language is spoken. On the other hand, it is still not clear how differences between languages may affect norms. The present study explores this issue by collecting and comparing norms on names and many other features from French Canadian speakers and English Canadian speakers living in Montreal, who thus live in similar cultural environments. Norms were collected for the photos of objects from the Bank of Standardized Stimuli (BOSS) by asking participants to name the objects, to categorize them, and to rate their familiarity, visual complexity, object agreement, viewpoint agreement, and manipulability. Names and ratings from the French speakers are available in Appendix A, available in the supplemental materials. The results show that most of the norms are comparable across linguistic groups and also that the ratings given are correlated across linguistic groups. The only significant group differences were found in viewpoint agreement and visual complexity. Overall, there was good concordance between the norms collected from French and English native speakers living in the same cultural setting.

Fast Wavelet-Based Image Characterization for Highly Adaptive Image Retrieval

Adaptive wavelet-based image characterizations have been proposed in previous works for content-based image retrieval (CBIR) applications. In these applications, the same wavelet basis was used to characterize each query image: This wavelet basis was tuned to maximize the retrieval performance in a training data set. We take it one step further in this paper: A different wavelet basis is used to characterize each query image. A regression function, which is tuned to maximize the retrieval performance in the training data set, is used to estimate the best wavelet filter, i.e., in terms of expected retrieval performance, for each query image. A simple image characterization, which is based on the standardized moments of the wavelet coefficient distributions, is presented. An algorithm is proposed to compute this image characterization almost instantly for every possible separable or nonseparable wavelet filter. Therefore, using a different wavelet basis for each query image does not considerably increase computation times. On the other hand, significant retrieval performance increases were obtained in a medical image data set, a texture data set, a face recognition data set, and an object picture data set. This additional flexibility in wavelet adaptation paves the way to relevance feedback on image characterization itself and not simply on the way image characterizations are combined.

Pattern recognition on a quantum computer

By means of a simple example it is demonstrated that the task of finding and identifying certain patterns in an otherwise (macroscopically) unstructured picture (data set) can be accomplished efficiently by a quantum computer. Employing the powerful tool of the quantum Fourier transform the proposed quantum algorithm exhibits an exponential speed-up in comparison with its classical counterpart. The digital representation also results in a significantly higher accuracy than the method of optical filtering. PACS: 03.67.Lx, 03.67.-a, 42.30.Sy, 89.70.+c.