A Brief Survey on Parallel Programming Applications for Image Processing Leticia Flores-Pulido, Esther Ortega-Mej´ıa, Eduardo Vega-Alvarado, Juan Manuel- ´ Alvarez Universidad Aut´ onoma de Tlaxcala Facultad de Ciencias B´ asicas, Ingenier´ıa y Tecnolog´ıa Calzada Apizaquito s/n. C.P. 90300 Apizaco, Tlaxcala, Mexico aicitel.flores@gmail.com, eortega m@hotmail.com, evega@ipn.mx, jumaalmx@gmail.com Abstract—In this work a series of applications for parallel programming is exposed, all of them related to image processing. A comparison of related works is included, showing the features and scope of these applications. The third section of this paper presents a proposal based on hypercubes, as a mechanism for establishing arrays of processors for parallel computing in addition to the approaches exposed in the included works, with the objective of exposing another kind of solution for image processing in huge data collections. I. I NTRODUCTION Parallel computing has different approaches and techniques, and uses multiple resources and platforms to solve complex problems, in those cases when computer power is required. All this diversity is related with the type of problem to be solved and the quantity and classification of the data involved. Once that the parallel section of an algorithm is identified, usually there are several alternatives for implementing a solution that takes advantage of this parallelism. A. Automatic Road Extraction From Satellite Images Using Extended Kalman Filtering And Efficient Particle Filtering There is an extended necessity for using geospatial infor- mation, in order to make precise visualizations of roads and terrain, in areas where projects are to be developed, making this a complex problem to be solved. The analysis of this information is made by processing satellite images, and the platform required has a considerably complexity. There are different methods to solve this problem. Automatic tracking starts with an automatic seeding of a road segment that indicates the road centreline, and then the computer learns relevant information of the road, such as initial direction, step size, width, and so on. A Road Extraction Method is shown in the flowchart of the Figure 1. It begins with a given satellite image, with two filters applied after the automatic seeding stage. The reference profile and seed point are extracted automatically by an automatic seeding technique. The edge of the road is detected using Canny edge detection. The road network is tracked by a filter until it reaches a road junction or an obstacle, then it is passed to another filter to initialize the seed point of the road branch. Fig. 1. Workflow of Road Extraction Method. Two algorithms are addressed for automatic road extraction: Extended Kalman Filtering (EKF) and Particle Filtering(PF). EKF serves to identify a hidden vari- able that represents the degree of freedom in the state vector of the dynamic system, considering the distance along the road as time variable. PF is a model- based estimation simulation used for Bayesian models. Since it is necessary to know the current road segment for the propagation of dynamic models, a pointer is assigned to the destination road segment. All information of the current seg- ment of the road, such as directions and neighbours, is indexed in the database by means of the pointer. The main approximation of the filter is the Gaussian as- sumption about the conditional target state distribution given a mode sequence and observations. The efficient particle filter with 80 particles yields satisfactory simulation results [1]. Extended Kalman Filter has many applications in comput- ing and is generally used to minimize error in Non lineal Dy- namical Systems. As a future work a neural network processed