Українська   Русский
DonNTU   Masters' portal

Abstract

Contents

Introduction

Nowadays, modern science and technology are developing more and more rapidly. Every day more and more people join this development, and less to stay away from it. Today, virtually impossible to find anyone who did not have a particular electronic device, mobile phone or personal computer, or - which is not used different software or the Internet.

Due to the increasing number of electronic devices, and increasing the amount of information that these devices is processed, transmitted and stored. Since the information in the modern world many times more than a person can handle in my life, then this information should be organized in order that it would take less time searching and search results to meet the demands.

Most of the information is already publicly available and systematized. In search of such information makes the World Wide Web and developed search engines. But the bulk of this information is a text, and it is not difficult to systematize. At the same time, most of unstructured information stored in the audio format in the images and video files. If the volume of this kind of information was not so great, it can be processed manually, describing what is in these files, but because the amount of information – ineffective manual processing.

Even though difficulties in processing audio, video and image files in the modern information technology, there are many automated systems to systematize such information. Such systems are of sufficient quality processed image files and audio files, relying on a very large computing power and heavy computational algorithms. But even such a system is still poorly treated video information, particularly if the video stream is continuous and processing should be carried out in real time.

In this paper will be reviewed the method of active contours, as a method by which you can select the contour of objects in the images in the video stream. After receiving the contours of objects, they can be classified and systematized according to the files in them. This method is able to obtain high accuracy outline of the object in the image, and also copes well with the processing of the video stream, by the fact that it works with the video as a sequence of images from and does not analyze frames as individual image. As discussed, various modifications of the method and their features.

1.1 Theme urgency

Problems of automated data processing have become more acute when humanity passed in the age of information technology, and the speed and quality of processing of stored information is no longer correspond to the quantity of information received. In the initial stages lacked various elementary recognition algorithms since the volume of detection was small, due to the low quality of the starting material. But with the growth of the quality of the image, and increased amounts of information to process, resulting in the emergence of new methods of detection, which relied on a variety of new technologies and new algorithms.

One such algorithm was proposed in 1988 by Michael Kass, Andrew Witkin and Demetrius Terzopoulos [3]. The idea of the algorithm is the gradual adjustment of the initial closed curve to the contour of the object, so that would be best suited to the edge of the object.

At first, this method was used mainly for the detection of persons in the image, while the used one of the first modifications [2], called the method of active models. But, over time, the method began to be used in various fields, particularly in medicine to analyze the X-ray and other images, as well as began to appear various modifications of the method based on various features of the areas in which they were applied, and compensate for the deficiencies of the algorithm. One of the facts that have influenced the development of the method is that the computing power of computers has increased several times, and was the method can be applied not only to a flat image, as well as to the video streams, and 3D image.

Therefore, research in this area in order to study this method, its applicability, and to find ways of its optimization and modifications are very relevant.

1.2 Goal and tasks of the research

The object of investigation is that of automated separation of different objects in the image, and in particular in the video stream.

The subject of the study – the allocation of edge of objects in the image by the method of active contours and its modification, especially modifications and ways to optimize the algorithm.

The aim of the final work is to study the modifications of existing active contours and to modify the method for selecting objects in the video stream.

In operation, the need to solve the following tasks:

  1. a study of the method of active contours;
  2. consideration of various modifications of the method for the detection of different types of objects;
  3. development of a method to allocate the various objects in the image and its optimization for use in the video stream;
  4. a software implementation of the developed method.

2 Current status of the problem

Target seting out in this section is to analyze and study the method of active models and existing versions. To do this, we have been studied various information sources that describe methods for isolating objects in the image by the method of active contours and its modifications.

In the source [3] is the classic description of the method of active contours. The basic formulas describing the energy parameters which express the contour. Also, the explanation of the method for processing stereo images in the case of processing a sequence of images, which is the same desired object.

The source [5] proposed to use a rough selection of the object in the image rectangular contour, while being based at the selection of the object, in the method of active contours. This modification of the method proposed for use in selection of objects in the image, which is not important object contour accuracy of detail, and it is important to define the position and approximate size of the object. Such systems can serve traffic control system or any other system of tracking objects.

In the source [4] describes the algorithm of the method of active contours to highlight defects in the image taken through a microscope. A feature extraction algorithm in this paper is that you want to select the defects, highlighting the objects themselves are not images.

The source [8] describes the operation method of active contours to highlight the defects of the road. Since defects are rarely similar to each other on the circuit parameters, the authors argue in the direction that would emphasize the work method in the direction of greater influence of external energy and reduce the impact of the internal. The consequence of this algorithm is a modification of a large adaptability of the method to the differences defects.

The source [9], the operation method of active contours in the allocation of defects of metal castings. The authors propose a method to modify by adding extra energy circuit, which are based on the distance of the contour points from the center of the selected shape and the average brightness of the selected object. These modifications make it possible to avoid the vagueness that appear due to the nature of the problem application, which uses the method of selection of objects in the image.

In reference [12] described a modification of the method of active contours, which uses multiple loops at the beginning of the selection. A feature of the modification is that the original image is covered with a plurality of primary circuits are in the process of delineating objects may merge into one. This allows you to select all objects in the image, avoiding the problem of an initial approximation to the object.

In the source [1] describes the modification of the method of active contours, which proposes the implementation of the possibility to break one loop for a few. This modification allows one circuit divided into more paths, if occurs select multiple objects that are not adjacent to each other.

In the source [2] describes one of the most popular ways to modify the classic algorithm of active contour method called active models. A feature of this modification is that inner strength contour expressed the desire deformation contour points to predetermined positions (the original template approach). Initially, this method of selection of objects image has been proposed to separate entities.

The sources [6] and [7] describes the use of the method of active models [2], making slight modifications. Thus, in [6] it was suggested that trained classifiers and use it in security systems. In [7] it is proposed to choose a template approach from the database templates that will more accurately allocate an object in case of severe distortions.

After analyzing the current literature, it can be concluded that now, for the selection of objects in the image by active circuits using a modification of the method, the purpose of which is a universal allocation of the object in the image and the ability of its classification. And also the fact that the method of active circuits is becoming increasingly popular, which accounts for the continuous development of various new modifications in the algorithm of the method.

3 Analysis and description of the classical algorithm of active contours

The basis of the method of active contours is that the circuit even before outlining the desired object has a certain initial form and, due to the different conditions affecting it, changes its shape, okonturivaya object - deformed (Figure 1).

In this model, the task of finding the borders of the object is stated as a change in position of contour points to new, in which the functional E – energy, wich is reaches to a minimum. The behavior of the active circuit (Figure 1), and its properties are completely determined by its functional E (energy yielding). The energy of the circuit depends on its shape, size, contour and its position in the image. It is written as a sum of two functions: the internal energy and outer energies (Formula 1.1).

Formula 1.1

Where a i b – weights, Eint – internal energy to the point, and Eext – external energy.

Активный контур

Figure 1 – Active contour;
(animation: 12 frames, 4 cycles of repetition, size - 188x79, 45 KB)

Internal energy – the energy of the broken line outline. This parameter is responsible for regulating the shape of the contour. The internal energy minimizes the broken line outline.

The external energy is responsible for the discrepancy between the contour image. The outer loop tries to minimize the difference circuit and the boundaries of the property which is outlined, and the smaller the difference is – the less the value of the external energy.

3.1 Usability of the method of active contours

The method of active contours is an effective method of selecting objects in the image. Despite the high computational complexity (if necessary detailed selection object outline) in a classic representation algorithm embodiment, the use of this method is widely and many modifications algorithm avoids the large volumes of calculations which are required for the object outline.

Thus, the algorithm method of active contours can be effectively used to highlight the contours of different objects in the image. Due to the fact that the loop is always closed – this simplifies its subsequent analysis and allows the labeling of an object based on the obtained contour. And the fact that you can control the release of the circuit by controlling the parameters of the internal and external energies of the circuit – it is possible to adapt the method for allocation of objects with known parameters.

Conclusion

The paper analyzed the method of active contours, which is used to highlight objects in the image. The analysis of this method, the same analysis was carried out varying the literature, which describes various modifications of the method and causes with which they are associated.

The analysis of sources led to the conclusion that the classical variant of the method of active contours is now developing in two directions: the algorithms which are based on adjusting the parameters of energy and the most similar to the classical idea of the method and algorithms that evolve along the path of the method of active models.

You can also notice that there is a tendency in the development of the method in any direction narrow selection of objects that are known in advance of the desired parameters of the circuit. This indicates that there is still no universal method of any allocation that would not require the initial information about the desired object.

Thus, the method of active circuits with proper optimization and modifications can be used to isolate any objects in the image or image sequence.

References

  1. Choi W.P., Lam K.M., Siu W.C. An adaptive active contour model for highly irregular boundaries //Pattern Recognition. – 2001. – Т. 34. – №. 2. – С. 323-331.
  2. Активные модели внешнего вида [Электронный ресурс]. – Режим доступа :http://habrahabr.ru/post/155759/.
  3. Kass M., Witkin A., Terzopoulos D. Snakes: Active contour models //International journal of computer vision. – 1988. – Т. 1. – №. 4. – С. 321-331.
  4. Бедзір А.О., Лютак І.З. Автоматичне знаходження контурів дефектів шляхом аналізу зображень, отриманих ультразвуковими методами контролю//Методи та приклади контролю якості. – 2009. – №. 22. – С. 28-32.
  5. Филатов В.Н., Смирнов В.М. Формирование адаптивного строба в видеосистемах наблюдения //Информационно-управляющие системы. – 2006. – №. 3.
  6. ГРЕЧУХИН И.А., КАРКИЩЕНКО А.Н. СТАТИСТИЧЕСКОЕ РАСПОЗНАВАНИЕ ЛИЦ ПО ГЕОМЕТРИИ ХАРАКТЕРНЫХ ТОЧЕК ДЛЯ СИСТЕМ ТРАНСПОРТНОЙ БЕЗОПАСНОСТИ //Управление большими системами: сборник трудов. – 2012. – №. 38.
  7. Бакина И., Местецкий Л., Цискаридзе А. Метод активного скелета в задаче распознавания формы изображений //Труды. – Т. 19. – С. 279-283.
  8. Судаков С. и др. Алгоритмы детектирования разметки и дефектов дорожного покрытия //Труды конференции GraphiCon. – 2008.
  9. Петров В.О., Привалов О.О. Модификация алгоритма активных контуров для решения задачи интерактивной сегментации растровых изображений дефектов металлических отливок //Научный журнал" Современные проблемы науки и образования. – 2008. – №. 6. – С. 32-41.
  10. Каркищенко А.Н. и др. Статистическое распознавание лиц по геометрии характерных точек для систем транспортной безопасности //М.: ИПУ РАН. – 2012. – С. 78-90.
  11. Запрягаев С.А. Программная оболочка для поиска примитивов на изображении / С.А. Запрягаев, А.И. Сорокин // Вестник ВГУ, серия: системный анализ и информационные технологии. – Воронеж, 2008. – № 2. – С. 37-47.
  12. A simple introduction of active contour with out ages [Электронный ресурс]. – Режим доступа:https://sites.google.com/site/rexstribeofimageprocessing/chan-vese-active-contours/wubiaotitiezi

Important note

In time of writing this abstract master's work is not yet complete. Estimated completion date - December 2015. Full text of the work and materials on the topic can be obtained from the author or his supervisor after that date.