In this work an adaptation of the Cross Industry Standard Process for Data Mining (CRISP-DM) methodology, in the context of digital medical image processing is proposed. Specifically, synthetic images reported in the literature are used as numerical phantoms. Construction of the synthetic images was inspired by a detailed analysis of some of the imperfections found in the real multilayer cardiac computed tomography images. Of all the imperfections considered, only Poisson noise was selected and incorporated into a synthetic database. An example is presented in which images contaminated with Poisson noise are processed and then subject to two classical digital smoothing techniques, identified as Gaussian filter and anisotropic diffusion filter. Additionally, the peak of the signal-to-noise ratio (PSNR) is considered as a metric to analyze the performance of these filters.

This paper presents the evaluation of two computational techniques for smoothing noise that might be present in synthetic images or numerical phantoms of magnetic resonance (MRI). The images that will serve as the data-bases (DB) during the course of this evaluation are available freely on the Internet and are reported in specialized literature as synthetic images called BrainWeb. The images that belong to this DB were contaminated with Rician noise, this being the most frequent type of noise in real MRI images. Also, the techniques that are usually considered to minimize the impact of Rician noise on the quality of BrainWeb images are matched with the Gaussian filter (GF) and an anisotropic diffusion filter, based on the gradient of the image (GADF). Each of these filters has 2 parameters that control their operation and, therefore, undergo a rigorous tuning process to identify the optimal values that guarantee the best performance of both the GF and the GADF. The peak of the signal-to-noise ratio (PSNR) and the computation time are considered as key elements to analyze the behavior of each of the filtering techniques applied. The results indicate that: a) both filters generate PSNR values comparable to each other. b) The GF requires a significantly shorter computation time to soften the Rician noise present in the considered DB.

Miguel Vera MSc, PhD1,2, https://orcid.org/0000-0001-7167-6356, Yoleidy Huérfano MSc2, https://orcid.org/0000-0003-0415-6654, Luis Javier Martínez PhD3, https://orcid. org/0000-0003-0917-9847, Yudith Contreras MSc1, https://orcid.org/0000-0003-4358-730X, Williams Salazar MD4, https://orcid.org/0000-0001-5669-6105, María Isabel Vera BSc4, https://orcid.org/0000-0003-1135-6283, Oscar Valbuena MSc5, https://orcid.org/0000-0003-3080-8839, Maryury Borrero MSc1, https://orcid.org/0000-00033025-1321, Carlos Hernández MSc1, https://orcid.org/0000-0001-8906-1982, Doris Barrera MSc1, https://orcid.org/0000-0002-6443-6757, Ángel Valentín Molina MSc3, https://orcid.org/0000-0001-9604-7222, Juan Salazar MSc1, https://orcid.org/0000-0001-6826-203X, Elkin Gelvez MSc1, https://orcid.org/0000-0001-5157-3341, Frank Sáenz MSc6, https://orcid.org/0000-0001-9604-7220 , Diego Hoyos BSc6, https://orcid.org/0000-0002-3341-2760, Yeny Arias BSc6, https://orcid.org/0000-0001-5574-4507. 1Universidad Simón Bolívar, Facultad de Ciencias Básicas y Biomédicas, Cúcuta, Colombia. *E-mail de correspondencia: m.avera@unisimonbolivar.edu.co 2Grupo de Investigación en Procesamiento Computacional de Datos (GIPCD-ULA), Universidad de Los Andes-Táchira, Venezuela. 3Grupo de Investigación en Ingeniería Clínica HUS (GINIC-HUS), Vicerrectoría de Investigación, Universidad ECCI. 4Servicio de Neurología, Hospital Central de San Cristóbal-Táchira, Venezuela. 5Grupo de Investigación en Educación Matemática, Matemática y Estadística (EDUMATEST), Facultad de Ciencias Básicas, Universidad de Pamplona. 6Universidad Simón Bolívar, Facultad de Ingeniería, Cúcuta, Colombia.

Este articulo propone una tecnica computacional no lineal para la segmentacion de los hematomas epidurales (EDH), presente en 7 bases de datos de imagenes cerebrales de tomografia multicapa. Esta tecnica consta de 3 etapas desarrolladas en el dominio tridimensional, a saber: preprocesamiento, segmentacion y cuantificacion del volumen ocupado por cada uno de los EDH segmentados. Para hacer juicios de valor sobre el rendimiento de la tecnica propuesta, las segmentaciones dilatadas de EDH, obtenidas automaticamente, y las segmentaciones de EDH, generadas manualmente por un neurocirujano, se comparan utilizando el coeficiente de Dice (Dc). La combinacion de parametros vinculados al valor mas alto de Dc define los parametros optimos de cada uno de los algoritmos computacionales que conforman la tecnica no lineal propuesta. Los resultados obtenidos permiten el reporte de un Dc superior a 0.90 que indica una buena correlacion entre las segmentaciones manuales y las producidas por la tecnica computacional desarrollada. Finalmente, como aplicacion clinica inmediata, considerando las segmentaciones automaticas, el volumen de cada hematoma se calcula considerando tanto el tamano del voxel de cada base de datos como el numero de voxeles que conforman los hematomas segmentados

This work evaluates the performance of some methods employed for assessing the volume of seven subdural hematomas (EDH), present in multi-layer computed tomography images. Firstly, a reference volume is considered to be that obtained by a neurosurgeon using the manual planimetric method (MPM). Secondly, the volume of the 7 EDHs is obtained considering both the original version of the ABC/2 method and two of its variants, identified in this paper as ABC/3 method and 2ABC/3 method. The ABC methods allow for calculation of the volume of the hematoma under the assumption that the EDH has an ellipsoidal shape. In third place, an intelligent automatic technique (SAT) is implemented that generates the three-dimensional segmentation of each EDH and from it the volume of the hematoma is calculated. The SAT consists of the pre-processing, segmentation and post-processing stages. In order to make judgments about the performance of the SAT, the Dice coefficient (Dc) is used to compare the dilated segmentations of the EDH with the EDH segmentations generated manually. Finally, the percentage relative error is calculated as a metric to evaluate the methodologies considered. The results show that the SAT method exhibits the best performance generating an average percentage error of less than 2%.

We present Jump, a practical system for capturing high resolution, omnidirectional stereo (ODS) video suitable for wide scale consumption in currently available virtual reality (VR) headsets. Our system consists of a video camera built using off-the-shelf components and a fully automatic stitching pipeline capable of capturing video content in the ODS format. We have discovered and analyzed the distortions inherent to ODS when used for VR display as well as those introduced by our capture method and show that they are small enough to make this approach suitable for capturing a wide variety of scenes. Our stitching algorithm produces robust results by reducing the problem to one of pairwise image interpolation followed by compositing. We introduce novel optical flow and compositing methods designed specifically for this task. Our algorithm is temporally coherent and efficient, is currently running at scale on a distributed computing platform, and is capable of processing hours of footage each day.

The deployment of pervasive displays in classrooms of children with severe autism is challenging. In this article, the authors explore the use of pervasive displays in special-education classrooms to help children with autism better reflect on their behaviors. They designed and developed three pervasive displays, each one varying its visualization in relation to targeted behaviors and the reinforcement mechanism used. BxColor mimics traditional practices by varying the color of "tags." BxPuzzle reinforces positive behavior by varying the clarity of puzzle pieces. BxBalloons penalizes negative behavior by deflating virtual aircrafts piloted by children. Each display was deployed in one classroom of children with severe autism for three weeks. The results indicate that BxColor was too abstract to be understood by participants. In contrast, both BxPuzzle and BxBalloons were instrumental in increasing behavior awareness, triggering social interactions, and promoting teamwork. This article is part of a special issue on pervasive displays.

While prior depth from focus and defocus techniques operated on laboratory scenes, we introduce the first depth from focus (DfF) method capable of handling images from mobile phones and other hand-held cameras. Achieving this goal requires solving a novel uncalibrated DfF problem and aligning the frames to account for scene parallax. Our approach is demonstrated on a range of challenging cases and produces high quality results.

Given a stereo pair it is possible to recover a depth map and use that depth to render a synthetically defocused image. Though stereo algorithms are well-studied, rarely are those algorithms considered solely in the context of producing these defocused renderings. In this paper we present a technique for efficiently producing disparity maps using a novel optimization framework in which inference is performed in “bilateral-space”. Our approach produces higher-quality “defocus” results than other stereo algorithms while also being 10 - 100× faster than comparable techniques.

This paper leverages occluding contours (aka “internal silhouettes”) to improve the performance of multi-view stereo methods. The contributions are 1) a new technique to identify free-space regions arising from occluding contours, and 2) a new approach for incorporating the resulting free-space constraints into Poisson surface reconstruction [14]. The proposed approach outperforms state of the art MVS techniques for challenging Internet datasets, yielding dramatic quality improvements both around object contours and in surface detail.

We address the problem of extending the field of view of a photo— an operation we call uncrop. Given a reference photograph to be uncropped, our approach selects, reprojects, and composites a subset of Internet imagery taken near the reference into a larger image around the reference using the underlying scene geometry. The proposed Markov Random Field based approach is capable of handling large Internet photo collections with arbitrary viewpoints, dramatic appearance variation, and complicated scene layout. We show results that are visually compelling on a wide range of real-world landmarks.

This paper leverages occluding contours (aka "internal silhouettes") to improve the performance of multi-view stereo methods. The contributions are 1) a new technique to identify free-space regions arising from occluding contours, and 2) a new approach for incorporating the resulting free-space constraints into Poisson surface reconstruction. The proposed approach outperforms state of the art MVS techniques for challenging Internet datasets, yielding dramatic quality improvements both around object contours and in surface detail.

This tutorial presents a hands-on view of the field of multi-view stereo with a focus on practical algorithms. Multi-view stereo algorithms are able to construct highly detailed 3D models from images alone. They take a possibly very large set of images and construct a 3D plausible geometry that explains the images under some reasonable assumptions, the most important being scene rigidity. The tutorial frames the multiview stereo problem as an image/geometry consistency optimization problem. It describes in detail its main two ingredients: robust implementations of photometric consistency measures, and efficient optimization algorithms. It then presents how these main ingredients are used by some of the most successful algorithms, applied into real applications, and deployed as products in the industry. Finally it describes more advanced approaches exploiting domain-specific knowledge such as structural priors, and gives an overview of the remaining challenges and future research directions.

This paper addresses the problem of obtaining 3d detailed reconstructions of human faces in real-time and with inexpensive hardware. We present an algorithm based on a monocular multi-spectral photometric-stereo setup. This system is known to capture high-detailed deforming 3d surfaces at high frame rates and without having to use any expensive hardware or synchronized light stage. However, the main challenge of such a setup is the calibration stage, which depends on the lights setup and how they interact with the specific material being captured, in this case, human faces. For this purpose we develop a self-calibration technique where the person being captured is asked to perform a rigid motion in front of the camera, maintaining a neutral expression. Rigidity constrains are then used to compute the head's motion with a structure-from-motion algorithm. Once the motion is obtained, a multi-view stereo algorithm reconstructs a coarse 3d model of the face. This coarse model is then used to estimate the lighting parameters with a stratified approach: In the first step we use a RANSAC search to identify purely diffuse points on the face and to simultaneously estimate this diffuse reflectance model. In the second step we apply non-linear optimization to fit a non-Lambertian reflectance model to the outliers of the previous step. The calibration procedure is validated with synthetic and real data.