Active Contour

Presentation of Snake (or active contour), and of the Greedy Algorithm.

Overview 

Active contours, called snake, is used to detect image's contours. It is based on several energies, externs, interns and contexts energies, which make snake move into images and make it stop if it detect a contour. The Greedy algorithm we used, is splited on five ponderated enregies : curve, intensity, continuity, gradient and balloon energy. The move is determined comparing differents fonctionnels of energy between all neighborhood of pixels of each point in a snake. In the Greedy algorithm alone, the ponderation of energies is static and we decided to change this and make it dynamic, according to influence of each energie. The first step was determine which energie was the more chosen. Then, we modify the coefficient of this more followed energie according each point for the local method and according all of points for the global method. Moreover, we modify coefficients in two differents ways : either we add a constant percentage (ex 10 %), or we add a percentage depending of the variation of energy. We check our algorithm in differents ways, with two kinds of images (real and artificial), with several percents of augmentation, and especially with two kinds of actives contours : the active contour with constants coefficients, and with the Autonomous Snake. 

Authors

Julien Michot - 2004

Download

• Report (french)

Image Compression Format

Dossier sur les formats de compression d'image (RLE, RL, HUFFMAN, LZW, JPEG, fractales et ondelettes).

Overview

En informatique, une image du type bitmap est représentée par un tableau en deux dimensions, dont les valeurs contenues dans les cases (ou pixel) caractérisent la couleur du pixel. Tandis qu'une image du type vectorielle est représentée par des équations mathématiques du style z=f(x,y), avec z la couleur et x,y les composantes spatiales. Il existe plusieurs codages pour symboliser la couleur, le plus connu étant le RGB (ou RVB : Rouge, Vert, Bleu), qui code une couleur selon ses trois composantes rouge, bleu et vert, ou le codage YUV utilisé par les formats PAL et SECAM et qui code la couleur selon la luminance et la chrominance. On distingue deux types de compression, la compression non destructrice qui, ne modifie en rien la qualité de l'image, et la compression dite destructrice qui elle élimine les plus insignifiantes données, comme notamment un changement rapide de chrominance entre deux pixels. Certains algorithmes (comme la compression de Huffman et LZW) réalisent un arbre ou un dictionnaire afin d'y ranger les redondances les plus fréquentes dans l'image. D'autres se chargent de repérer les formes qui se reproduisent comme le codage par fractale ou RLE. Afin, un autre format, le JPEG, découpe l'image en bloc de 8x8 pixels pour le JPEG et en lignes pour le JPEG 2000, puis réalise une transformation mathématique et change la variation de couleur en somme de fonctions sinusoïdales pour l'un et en somme d'ondelettes pour l'autre. 

Authors

Julien Michot - 2003

Download

• Report (french)

Codages de l'information en informatique industrielle

Dossier sur les codages d'une information binaire en Infomatique Industrielle (RZ, NRZ, NRZI, MANCHESTER, MANCHESTER différentiel, Miller, CMI, MLT3, Bipolaire simple et entrelacé, HDBn, ainsi que les codes à transformation de Valence

Overview

Coding is the operation witch realizes the correspondence between a logical state and a physical value. The applications of the industrial data processing, in Internet and other networks, required the use of new methods of coding adapted to this various employment and their own needs. There are two principal categories of coding, the binary codes and the codes on multiple levels. But there are two sub-groups in those categories, which not use the same kind of coding. One's, the coding by levels, makes correspond in a logical state, a level of voltage whereas the other, the coding by transition, uses a jump of voltage to code data. The transition is used to reduce the error rate at the time of the reception. Moreover, certain codes use the output signal to remake the clock. A lot of problems can deteriorate the quality of a code, they are mainly induced by the characteristics of the physical support. An electric signal, current, frequency, and voltage can be parasitized, attenuated, or filtered by his surroundings. With a coding by level of voltage, the use of the null voltage is doubtful for the receiver because the tension received at the time of a loss of transmission, is null too. Moreover, when a continuation from several 0 or 1 occurs, the receiver cannot distinguish the clock into the signal. However the method of coding by transition reduce the maximum flow, so it's not better. The risk of inversion of wire was also studied, and others codings known as differential were conceived. The purpose of the codes with transformation of valence is coding several bits by front of clock. 

Authors

Julien Michot - 2003

Download

• Report (french)

CD-ROM & CR-R

Report on the CD-ROM and CD-R.

Overview

CD-ROM is a derivative of the "Compact Disc", and makes it possible to store many numerical information. This support is composed of a flat disc 12 cm in diameter, in its most widespread version, and consists of several layers, plastics and metal. To code binary data, the metal layer is used. This part of CD-ROM to the property to return a laser beam if one sends a laser on a plane part. On the other hand, when a hollow is crossed, an cell of 167 Nm, the laser beam is not reflected any more, and is cancelled. Thus, to code binary data, one uses hollows and dishes. One 1 will be to code by a transition, i.e. the passage of a hollow to a dish and the reverse. 0 logic will be him to code by a hollow or a dish, according to alternation. For technical reasons (precision of engraving, safety of the reading), it is necessary to have at least two bits with 0 between two 1 consecutive, and to the maximum 10 consecutive bits with zero between two bits with 1. This system of coding appèle the EFM (Eight to Fourteen Modulation). The logic design of CD-Romanian is made up of three parts. "Lead-in", in the center of CD, which contains formations describing the contents of the support, hierarchy... The zone of program contains all the files, all the binary data of the applications. Lastly, there is the zone of "Lead-out" which marks the end of CD-ROM. 

Authors

Julien Michot - 2003

Download

• Report (french)