Juan Domingo

Applying logistic regression to relevance feedback in image retrieval systems

This paper deals with the problem of image retrieval from large image databases. A particularly interesting problem is the retrieval of all images which are similar to one in the users mind, taking into account his/her feedback which is expressed as positive or negative preferences for the images that the system progressively shows during the search. Here we present a novel algorithm for the incorporation of user preferences in an image retrieval system based exclusively on the visual content of the image, which is stored as a vector of low-level features. The algorithm considers the probability of an image belonging to the set of those sought by the user, and models the logit of this probability as the output of a generalized linear model whose inputs are the low-level image features. The image database is ranked by the output of the model and shown to the user, who selects a few positive and negative samples, repeating the process in an iterative way until he/she is satisfied. The problem of the small sample size with respect to the number of features is solved by adjusting several partial generalized linear models and combining their relevance probabilities by means of an ordered averaged weighted operator. Experiments were made with 40 users and they exhibited good performance in finding a target image (4 iterations on average) in a database of about 4700 images. The mean number of positive and negative examples is of 4 and 6 per iteration. A clustering of users into sets also shows consistent patterns of behavior.

Spatial size distributions: applications to shape and texture analysis

This paper proposes new descriptors for binary and gray-scale images based on newly defined spatial size distributions (SSD). The main idea consists of combining a granulometric analysis of the image with a comparison between the geometric covariograms for binary images or the auto-correlation function for gray-scale images of the original image and its granulometric transformation; the usual granulometric size distribution then arises as a particular case of this formulation. Examples are given to show that in those cases in which a finer description of the image is required, the more complex descriptors generated from the SSD could be advantageously used. It is also shown that the new descriptors are probability distributions so their intuitive interpretation and properties can be appropriately studied from the probabilistic point of view. The usefulness of these descriptors in shape analysis is illustrated by some synthetic examples and their use in texture analysis is studied. Various cases of SSD and several former methods for texture classification are compared.

A novel Bayesian framework for relevance feedback in image content-based retrieval systems

This paper presents a new algorithm for image retrieval in content-based image retrieval systems. The objective of these systems is to get the images which are as similar as possible to a user query from those contained in the global image database without using textual annotations attached to the images. The main problem in obtaining a robust and effective retrieval is the gap between the low level descriptors that can be automatically extracted from the images and the user intention. The algorithm proposed here to address this problem is based on the modeling of user preferences as a probability distribution on the image space. Following a Bayesian methodology, this distribution is the prior distribution and its parameters are modified based on the information provided by the user. This yields the a posteriori from which the predictive distribution is calculated and used to show to the user a new set of images until he/she is satisfied or the target image has been found. Experimental results are shown to evaluate the method on a large image database in terms of precision and recall.

Combining similarity measures in content-based image retrieval

The purpose of content based image retrieval (CBIR) systems is to allow users to retrieve pictures from large image repositories. In a CBIR system, an image is usually represented as a set of low level descriptors from which a series of underlying similarity or distance functions are used to conveniently drive the different types of queries. Recent work deals with combination of distances or scores from different and usually independent representations in an attempt to induce high level semantics from the low level descriptors of the images. Choosing the best method to combine these results requires a careful analysis and, in most cases, the use of ad-hoc strategies. Combination based on or derived from product and sum rules are common approaches. In this paper we propose a method to combine a given set of dissimilarity functions. For each similarity function, a probability distribution is built. Assuming statistical independence, these are used to design a new similarity measure which combines the results obtained with each independent function.

A naive relevance feedback model for content-based image retrieval using multiple similarity measures

This paper presents a novel probabilistic framework to process multiple sample queries in content based image retrieval (CBIR). This framework is independent from the underlying distance or (dis)similarity measures which support the retrieval system, and only assumes mutual independence among their outcomes. The proposed framework gives rise to a relevance feedback mechanism in which positive and negative data are combined in order to optimally retrieve images according to the available information. A particular setting in which users interactively supply feedback and iteratively retrieve images is set both to model the system and to perform some objective performance measures. Several repositories using different image descriptors and corresponding similarity measures have been considered for benchmarking purposes. The results have been compared to those obtained with other representative strategies, suggesting that a significant improvement in performance can be obtained.

Irregular motion recovery in fluorescein angiograms

Abstract Fluorescein angiography is a common procedure in ophthalmic practice, mainly to evaluate vascular retinopathies and choroidopathies from sequences of ocular fundus images. In order to compare the images, a reliable overlying is essential. This paper proposes some methods for the recovery of irregular motion in fluorescein angiograms (FA). The overlying is done by a three step procedure: detection of relevant points, matching points from different images and estimation of the assumed linear geometric transformation. A stochastic model (closely related to the general linear model) allows to fuse the second and third steps. Two different estimators of the geometric transformation are proposed and tested with real FAs. Images from choroido-retinal diseases have been analysed: diabetic retinopathy, vein occlusions and choroidal neovascular membrane. Results have been evaluated using a different number of relevant points with different spatial arrangements. Registration accuracy is evaluated as the mean squared error between real and transformed relevant point locations for those points not used to estimate the transformation.

On the advantages of combining differential algorithms and log-polar vision for detection of self-motion from a mobile robot

Abstract This paper describes the design and implementation on programmable hardware (FPGAs) of an algorithm for the detection of self-mobile objects as seen from a mobile robot. In this context, ‘self-mobile’ refers to those objects that change in the image plane due to their own movement, and not to the movement of the camera on board of the mobile robot. The method consists on adapting the original algorithm from Chen and Nandhakumar [A simple scheme for motion boundary detection, in: Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, 1994] by using foveal images obtained with a special camera whose optical axis points towards the direction of advance. It is shown that the use of log-polar geometry simplifies the original formulation and highly reduces the volume of data to be treated. Limitations of the algorithm due to the differential nature of the approach are discussed, relating them with the parameters of the system. Experiments are shown in which a self-mobile object is detected in several conditions.

Selecting the structuring element for morphological texture classification

This paper deals with a concrete aspect of texture classification: the choice of a good structuring element (SE) when the texture features used for classification are obtained from morphological granulometries. First, a granulometry is defined from the morphological opening of the texture using a convex and compact subset containing the origin as SE. Then, some usual distributional descriptors (mean, variance, skewness and kurtosis) of the granulometric size distribution are used as texture features. The main point of the paper is the choice of a good SE from the point of view of texture classification. A methodology is explained and software has been developed that helps in such a choice, for any given criterion for the quality of the classification.

Effect of a Diet Supplemented with α-Tocopherol and β-Carotene on ATP and Antioxidant Levels after Hepatic Ischemia-Reperfusion

Ischemia-reperfusion injury associated with liver transplantation remains a serious complication in clinical practice. In the present study the effect of intake of α-tocopherol or β-carotene to limit liver injury by oxidative stress in ischemia and reperfusion was explored. Wistar rats were fed with diets enriched with α-tocopherol (20 mg/day) or β-carotene (3 mg/day) for 21 days. After 21 days, their livers were subjected to 15 and 30 min of ischemia and afterwards were reperfused for 60 min. The recovery of levels of ATP during reperfusion was better in the group of rats whose diets were supplemented with α-tocopherol or β-carotene than in the group control. The supplementation of the diet induced changes in the profile of enzymatic antioxidants. The supplementation with α-tocopherol and β-carotene resulted in a decreased of superoxide dismutase during the ischemia and a recovery was observed after reperfusion. Not changes were observed for the enzymes catalase and glutathione peroxidase and glutathione but their values were higher to those of the group control. In conclusion, the supplementation with α-tocopherol and β-carotene improve the antioxidant and energetic state of liver after ischemia and reperfusion injury.

Apparel sizing using trimmed PAM and OWA operators

This paper is concerned with apparel sizing system design. One of the most important issues in the apparel development process is to define a sizing system that provides a good fit to the majority of the population. A sizing system classifies a specific population into homogeneous subgroups based on some key body dimensions. Standard sizing systems range linearly from very small to very large. However, anthropometric measures do not grow linearly with size, so they can not accommodate all body types. It is important to determine each class in the sizing system based on a real prototype that is as representative as possible of each class. In this paper we propose a methodology to develop an efficient apparel sizing system based on clustering techniques jointly with OWA operators. Our approach is a natural extension and improvement of the methodology proposed by McCulloch, Paal, and Ashdown (1998), and we apply it to the anthropometric database obtained from a anthropometric survey of the Spanish female population, performed during 2006.