Search

Retinal images are used for diagnostic purposes by ophthalmologists. However, despite controlled conditions in acquisition retinal images often suffer from non-uniform illumination which hinder their clinical use. In this work we propose the compensation of the illumination by modeling the intensity as a sum of non-stationary signals using bidimensional empirical mode decomposition (BEMD). We compare the estimation and compensation of the background illumination with a widely used technique based retinal image pixel classification. The proposed method has shown to provide a better estimation of the background illumination, particularly in complicated areas such as the optic disk (usually bright) and the periphery of fundus images (usually dim). © 2015 SPIE. Prometeo - Secretaria de Educacion Superior, Ciencia, Tecnologia e Innovacion;Red Nacional de Investigacion y Educacion del Ecuador;Universidad Nacional de Colombia Sede Bogota, CIM at lab

Retinal vessel segmentation is a prerequisite for the analysis of vessel parameters such as tortuosity, variation of the vessel width along the vessel and the ratio between the venous and arterial vessel width. This analysis can provide indicators for the presence of a wide range of diseases. Different types of approaches have been proposed to segment the retinal vasculature and two important groups are vessel tracking and pixel processing based methods. An advantage of tracking based methods is the guaranteed connectedness of vessel segments, in pixel processing based methods connectedness is not guaranteed. In this work an automated vessel linking framework is presented. The framework links together separate pieces of the retinal vasculature into a connected vascular tree. To determine which vessel sections should be linked together the use of a supervised cost function is proposed. Evaluation is performed on the vessel centerlines. The results show that the vessel linking framework outperforms other automated vessel linking methods especially for the narrowest vessels.

As a result of the advances in skin imaging technology and the development of suitable image processing techniques, during the last decade, there has been a significant increase of interest in the computer-aided diagnosis of skin cancer. Dermoscopy is a non-invasive skin imaging technique which permits visualization of features of pigmented melanocytic neoplasms that are not discernable by examination with the naked eye. One of the useful features in dermoscopic diagnosis is the blue-white veil (irregular, structureless areas of confluent blue pigmentation with an overlying white "ground-glass" film) which is mostly associated with invasive melanoma. In this preliminary study, a machine learning approach to the detection of blue-white veil areas in dermoscopy images is presented. The method involves pixel classification based on relative and absolute color features using a decision tree classifier. Promising results were obtained on a set of 224 dermoscopy images.

We present Genie Pro, a new software tool for image analysis produced by the ISIS (Intelligent Search in Images and Signals) group at Los Alamos National Laboratory. Like the earlier GENIE tool produced by the same group, Genie Pro is a general purpose adaptive tool that derives automatic pixel classification algorithms for satellite/aerial imagery, from training input provided by a human expert. Genie Pro is a complete rewrite of our earlier work that incorporates many new ideas and concepts. In particular, the new software integrates spectral information; and spatial cues such as texture, local morphology and large-scale shape information; in a much more sophisticated way. In addition, attention has been paid to how the human expert interacts with the software: Genie Pro facilitates highly efficient training through an interactive and iterative “training dialog”. Finally, the new software runs on both Linux and Windows platforms, increasing its versatility. We give detailed descriptions of the new techniques and ideas in Genie Pro, and summarize the results of a recent evaluation of the software.

In this work we compare the performance of a number of vessel segmentation algorithms on a newly constructed retinal vessel image database. Retinal vessel segmentation is important for the detection of numerous eye diseases and plays an important role in automatic retinal disease screening systems. A large number of methods for retinal vessel segmentation have been published, yet an evaluation of these methods on a common database of screening images has not been performed. To compare the performance of retinal vessel segmentation methods we have constructed a large database of retinal images. The database contains forty images in which the vessel trees have been manually segmented. For twenty of those forty images a second independent manual segmentation is available. This allows for a comparison between the performance of automatic methods and the performance of a human observer. The database is available to the research community. Interested researchers are encouraged to upload their segmentation results to our website (http://www.isi.uu.nl/Research/Databases). The performance of five different algorithms has been compared. Four of these methods have been implemented as described in the literature. The fifth pixel classification based method was developed specifically for the segmentation of retinal vessels and is the only supervised method in this test. We define the segmentation accuracy with respect to our gold standard as the performance measure. Results show that the pixel classification method performs best, but the second observer still performs significantly better.

Soccer is a very popular sport all over the world, Coaches and sport commentators need accurate information about soccer games, especially about the players behavior. These information can be gathered by inspectors who watch the soccer match and report manually the actions of the players involved in the principal phases of the game. Generally, these inspectors focus their attention on the few players standing near the ball and don't report about the motion of all the other players. So it seems desirable to design a system which automatically tracks all the players in real- time. That's why we propose to automatically track each player through the successive color images of the sequences acquired by a fixed color camera. Each player which is present in the image, is modelized by an active contour model or snake. When, during the soccer match, a player is hidden by another, the snakes which track these two players merge. So, it becomes impossible to track the players, except if the snakes are interactively re-initialized. Fortunately, in most cases, the two players don't belong to the same team. That is why we present an algorithm which recognizes the teams of the players by pixels representing the soccer ground which must be withdrawn before considering the players themselves. To eliminate these pixels, the color characteristics of the ground are determined interactively. In a second step, dealing with windows containing only one player of one team, the color features which yield the best discrimination between the two teams are selected. Thanks to these color features, the pixels associated to the players of the two teams form two separated clusters into a color space. In fact, there are many color representation systems and it's interesting to evaluate the features which provide the best separation between the two classes of pixels according to the players soccer suit. Finally, the classification process for image segmentation is based on the three most discriminating color features which define the coordinates of each pixel in an 'hybrid color space.' Thanks to this hybrid color representation, each pixel can be assigned to one of the two classes by a minimum distance classification.

Local granulometric size distributions are generated by performing a granulometry on an image and keeping local pixel counts in a neighborhood of each pixel at the completion of each successive opening. Normalization of the resulting size distributions yields a probability density at each pixel. These densities contain texture information local to each pixel. Pixels can be classified according to the moments of the densities. Further refinement can be accomplished by employing several structuring-element sequences in order to generate a number of granulometries, each revealing different texture qualities. Classification is accomplished by comparing observed moments to those representing a database of textures. The collection of database moments are actually random variables dependent on random texture processes, and the method employed in the present paper involves the comparison of observed moments to the means of database-texture moments.