Sérgio Francisco da Silva

Improving the ranking quality of medical image retrieval using a genetic feature selection method

In this paper, we take advantage of single-valued functions that evaluate rankings to develop a family of feature selection methods based on the genetic algorithm approach, tailored to improve the accuracy of content-based image retrieval systems. Experiments on three image datasets, comprising images of breast and lung nodules, showed that developing functions to evaluate the ranking quality allows improving retrieval performance. This approach produces significantly better results than those of other fitness function approaches, such as the traditional wrapper and than filter feature selection algorithms.

An image retrieval system adaptable to user's interests by the use of relevance feedback via genetic algorithm

The emergence of multimedia technology and the rapid expansion of image sets on the internet have attracted a lot of research tools for effective retrieval of visual data. When working in the image retrieval context the main goal is to retrieve images which might be useful or relevant to the user based on features automatically extracted from the images. The proposal of this work is to integrate the information provided by the user into the decision procedure by the use of the relevance feedback mechanism. The relevance feedback technique used is based on genetic algorithms using a proposed order-based fitness function in order to adapt the users image similarity criteria. Image similarity is expressed as a weighted integration of color, shape and texture features. The retrieval process itself is based on the Local Similarity Pattern, where the image areas are uniformly partitioned into regions, and the similarity between the images is measured by corresponding region similarities. The use of negative and positive weights for the features, into the genetic algorithm, allows one to express, in a continuous way, the concepts of relevance, irrelevance and undesirability in the similarity model used. Experiments in a database with 12750 images has shown that the integration of features through the proposed genetic algorithm into a relevance feedback mechanism provides good results in the image retrieval context.

Feature space optimization for content-based image retrieval

Substantial benefits can be gained from effective Relevance Feedback techniques in content-based image retrieval. However, existing techniques are limited due to computational cost and/or by being restricted to linear transformations on the data. In this study we analyze the role of nonlinear transformations in relevance feedback. We present two promising Relevance Feedback methods based on Genetic Algorithms used to enhance the performance on the task of image retrieval according to the users interests. The first method adjusts the dissimilarity function by using weighting functions while the second method redefines the features space by means of linear and nonlinear transformation functions. Experimental results on real data sets demonstrate that our methods are effective and the results show that the transformation approach outperforms the weighting approach, achieving a precision gain of up to 70%. Our results indicate that nonlinear transformations have a great potential in capturing the users interests in image retrieval and should be further analyzed employing other learning/optimization mechanisms.

Silhouette-based feature selection for classification of medical images

Classification is an important task for computer-aided diagnosis systems (CADs). However, many classifiers may not perform well, presenting poor generalization and high computational cost, especially when dealing with high-dimensional datasets. Thus, feature selection can greatly mitigate these problems. In this paper, we propose two filter-based feature selection algorithms that calculate the simplified silhouette statistic as evaluation function: the silhouette-based greedy search (SiGS) and the silhouette-based genetic algorithm search (SiGAS). Silhouette statistic is used to guide the search for features that provide better class separability. Experiments performed on three datasets have shown that the SiGAS algorithm overcomes traditional filter algorithms, such as CFS, FCBF and reliefF. It also outperforms a similar algorithm, kNNGAS, based on genetic algorithm that minimizes the classification error of k-nearest neighbors. Additionally, results have shown that SiGAS produces better accuracy than SiGS.

A Multi-Dimensional Similarity Modeling and Relevance Feedback Approach for Content-Based Image Retrieval.

This work presents a multi-dimensional similarity modeling strategy and relevance feedback technique for minimizing the semantic gap intrinsic problem of CBIR systems by allowing users to customize their queries according to their requirements and preferences. We propose a composite strategy using a multi-dimensional, vectorial, spatially clustered, and relevance-ordered approach. Given a set of k features which represents the images in an image database, the similarity measure between a query image and another from the image collection is analyzed in k components, and the images are ranked on a A dimensional space according to their projections over the axis x n , where n = 1,2,... k. System experimentation was executed thoroughly using a test image database containing up to 12,000 pictures. The experimental results have shown that the presented approach can substantially improve the outcome in image retrieval systems.

Images segmentation using a modified Hopfield artificial neural network

Good image segmentation can be achieved by finding the optimum solution to an appropriate energy function. A Hopfield neural network has been shown to solve complex optimization problems fast, but it only guarantees convergence to a local minimum of the optimization function. This paper proposed a little modification in the Hopfield neural network, the experimental results were made with proposed model for gray-scale images segmentation on synthetic and satellite images showing its effectiveness.

Findings on ranking evaluation functions for feature weighting in image retrieval

BackgroundThere are substantial benefits to be gained from ranking optimization in several information retrieval and recommendation systems. However, the analysis of ranking evaluation functions (REFs), which play a major role in many ranking optimization models, needs to be further investigated. An analysis of previous studies that investigated REFs was performed, and evidence was found which indicated that the choice of a proper REF is context sensitive.MethodsIn this study, we analyze a broad set of REFs for feature weighting aimed at increasing the image retrieval effectiveness. The REFs analyzed sums ten and includes the most successful and representative REFs from the literature. The REFs were embedded into a genetic algorithm (GA)-based relevance feedback (RF) model, called WLSP-C ±, aimed at improving image retrieval results through the use of learning weights for image descriptors and image regions.ResultsAnalyses of precision-recall curves in five real-world image data sets showed that one non-parameterized REF named F5, not analyzed in previous studies, overcame recommended ones, which require parameter adjustment. We also provided a computational analysis of the GA-based RF model investigated, and it was shown that it is linear in regard to the image data set cardinality.ConclusionsWe conclude that REF F5 should be investigated in other contexts and problem scenarios centered on ranking optimization, as ranking optimization techniques rely heavily on the ranking quality measure.

Image retrieval employing genetic dissimilarity weighting and feature space transformation functions

We present two promising Relevance Feedback methods based on Genetic Algorithms used to enhance the performance on the task of image retrieval according to the users interests. The first method adjusts the dissimilarity function by using weighting functions while the second method redefines the feature space by means of linear and nonlinear transformation functions. Experimental results on real datasets demonstrate that our methods are effective and the results show that the transformation approach outperforms the weighting approach, achieving a precision gain of up to 70%.

H-METRIC: CHARACTERIZING IMAGE DATASETS VIA HOMOGENIZATION BASED ON KNN-QUERIES

Precision-Recall is one of the main metrics for evaluating content-based image retrieval techniques. However, it does not provide an ample perception of the properties of an image dataset immersed in a metric space. In this work, we describe an alternative metric named H-Metric, which is determined along a sequence of controlled modifications in the image dataset. The process is named homogenization and works by altering the homogeneity characteristics of the classes of the images. The result is a process that measures how hard it is to deal with a set of images in respect to content-based retrieval, offering support in the task of analyzing configurations of distance functions and of features extractors.