EP1766579A2 - Procede de detection de structures geometriques dans des images - Google Patents

Procede de detection de structures geometriques dans des images

Info

Publication number
EP1766579A2
EP1766579A2 EP05769803A EP05769803A EP1766579A2 EP 1766579 A2 EP1766579 A2 EP 1766579A2 EP 05769803 A EP05769803 A EP 05769803A EP 05769803 A EP05769803 A EP 05769803A EP 1766579 A2 EP1766579 A2 EP 1766579A2
Authority
EP
European Patent Office
Prior art keywords
sector
transformation
boundary line
space
images
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP05769803A
Other languages
German (de)
English (en)
Inventor
Achim Kirsch
Fritz Jetzek
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
PerkinElmer Cellular Technologies Germany GmbH
Original Assignee
Evotec Technologies GmbH
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Evotec Technologies GmbH filed Critical Evotec Technologies GmbH
Priority to EP05769803A priority Critical patent/EP1766579A2/fr
Publication of EP1766579A2 publication Critical patent/EP1766579A2/fr
Withdrawn legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/12—Edge-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/48—Extraction of image or video features by mapping characteristic values of the pattern into a parameter space, e.g. Hough transformation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10056—Microscopic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20048—Transform domain processing
    • G06T2207/20061—Hough transform
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30024—Cell structures in vitro; Tissue sections in vitro

Definitions

  • the invention relates to a method for detecting geometric structures in images, in particular in images of chemical and / or biological samples, such as images of cells.
  • the invention relates to a method for the detection of cell traces.
  • maxima can arise in the transformed space that do not correspond to any object actually present in the Oitsraum.
  • the phenomenon occurs, for example, when the intensities recorded "randomly" during the transformation of a point are larger than those actually associated with the object ("correlated noise", Leavers a.a.O.).
  • Shpilman et al. describes in SHPILMAN, TL and V. URAILOVSKY: Fast and Robust Techniques for Detecting Straight Line Segments Usmg Locals Models. Pattern Recognition Letters, 20: 865-877, 1999 a method for the detection of straight lines, wi ⁇ ? D used in the existing knowledge of partial points of the objects to be recognized.
  • One-dimensional parameter spaces are used here.
  • the image to be analyzed is preprocessed with an edge filter, so that the algorithm only records points lying on lines from the outset. For each of these points p k , a one-dimensional histogram is generated which indicates the angle under which as many of the remaining points q k as possible have a line; ' form.
  • the method produces errors if preprocessing leaves too many points that are not on lines.
  • the authors state that such points, in the vicinity of /; can not be included in the analysis; the transformation algorithm is based on a point determination of the line pq t with a fixed reference line and will lead to distortions if the distance between p and q k is too short.
  • the possibility of falsely recognized lines continues to be observed (see Fig. 9 (b) in Shpilman aao). - ⁇ -
  • the Hough transformation has already been reported in the biological field by Lyazghi et al. ⁇ LYAZGHI, A., C. DECAESTEKER, I. CAMBY, R. KISS, and P V. HAM Chatactensation of Actin Filaments in Cancer Cells by the Hough Ti ans Form. Signal Processing, Pattern Recognition and Applications, pp. 138-142 , IASTED, July 2001.) used to detect fibers within the cytoskeleton of cancer cells.
  • the method also allows false maxima during the (integral) transformation of the cell surface. The extreme positions are only checked for their validity in a post-processing. In this case, the authors compare the length of the corresponding line with the expected length, with the cell diameter serving as the maximum extent.
  • the object of the invention is to provide a method for detecting geometric Struktui s in pictures, through which the geometric Stuctures can be detected with a high degree of security.
  • geometric structures are detected in images. These are, in particular, images of chemical and / or biological samples and particularly preferably images of cells, cell organelles and / or cell nuclei.
  • images are, in particular, images of chemical and / or biological samples and particularly preferably images of cells, cell organelles and / or cell nuclei.
  • a Detcktieren a Begrenzungshnie an imaged object.
  • several imaged objects are detected by the individual boundary patterns. This can be done, for example, by a threshold value method and corresponding arithmetic operations. Subsequently, a sector is defined within the image.
  • the sector is in particular circular or circular segment-shaped.
  • a sector in the form of a segment of a circle preferably extends outward in a funnel shape from a sector origin lying on the boundary line, which is in particular the center of the circle.
  • the image section defined by the sector is transformed into a transformation space.
  • a transformation is used by which the geometric structures present in the image section, in particular cell traces, respectively correspond to signatures in the transformation space.
  • the signatures are maxima in the transformation space.
  • the transformation space it is then determined whether one or more signatures exist.
  • the determined signatures are then transformed back from the transformation space so that the geometric structure is shown in the corresponding sector.
  • the method according to the invention has the advantage that a sector uniquely defined in its position is transformed, the sector space being defined in particular by the sector origin lying on the boundary line of the object.
  • the transformation is preferably carried out by a linear, circular or generalized Hough transformation.
  • the transformation is performed in such a way that the spatial structures of the budget detail are assigned signatures in the transformation space which are in a fixed spatial relationship to the sector origin, in particular intersecting the sector origin or - G -
  • the defined sector has an axis of symmetry, the orientation of the axis of symmetry being substantially perpendicular to the boundary line.
  • the sector extends outward from the boundary line of the object.
  • the image section is preferably an intersection of the sector and a zone extending along the boundary line, in particular outside the boundary line. In the context of the detection of cell traces, this zone is referred to as fiber space in particular.
  • the sectors are analyzed in common, taking into account the sequence of their respective sector origins on the boundary line.
  • the sequence of the contour pixels to be analyzed can be calculated according to the following classical algorithm.
  • the point is defined as the first element of the sequence to be determined and removed from the set of contour pixels.
  • FIG. 1 shows a schematic sketch of an imaged object in the form of a cell for explaining the method according to the invention
  • Fig. 2a - 2e Illustrations of a cell-stained cells in individual stages of the method according to the invention.
  • the image to be analyzed on cell traces is first preprocessed by marking the relevant areas.
  • the environmental points of the cells lie on the scale in between and can be marked, for example, by a simple segmentation with global thresholds.
  • the marking itself should then be in the form of a binary mask, in which the areas belonging to the individual cells, the fiber spaces, can be distinguished from one another.
  • the cell bodies themselves are not included in these regions, so that each fiber space has outer and inner boundary lines.
  • a circle sector is calculated, the center of which is> k and protrudes from the boundary line in the fiber circumference so that the tangent of the border at p k forms a right angle with the Richtraumsvektoi r of the sector.
  • the opening angle of the sector is defined by 2 ⁇ where a fiber protruding from the boundary is from a right angle!
  • the detection can also be limited to those fibers which can deviate only in one direction by ⁇ , but in the other direction by a second tolerance angle p. In this case, the opening angle should be ⁇ + p.
  • the circular sector is brought to intersection with fiber space. The length of r must be such that all points on the circumference are separated.
  • the labels contained in FIG. 1 are characterized as follows: a, b, c, d, f, p and r are vectors; F, H, R and K denote sets of vectors, A x and ⁇ ⁇ 'are scalar magnitudes. By ⁇ j>) the scalar product of the vectors a and b is denoted; p and ⁇ are angles; ⁇ r [denotes the Euclidean length of r, and denoted by (- «, ⁇ «) 'is a transposed vector whose components use the X or Y portions of the vector a.
  • Fig. 1 shows a circular sector R with the center p ⁇ , on the surrounding line of a cell K.
  • the orientation of the sector is given by the vector r and the angles p and ⁇ .
  • the sector intersects fiber space F and background H.
  • the local orientation of the sector with respect to the circular line can be determined unambiguously in the continuous space via the local tangent slope. However, if the cell boundary is present as a chain of points in a discrete grid, the orientation can only be approximated.
  • the problem was eg Utcke [Utcke, S.: Error-Bounds on Curvature Estinnation. In: Griffin, L. and M. Lillholm (ed.): Scale-Space 2003, pp. 657-666. Springer-Verlag, 2003].
  • One known method for calculating local slope in discrete space is the use of an isosceles triangle that is moved along the contour line to be analyzed.
  • the legs of the same length are indicated in Figure 1 by the vectors a and b. Their endpoints lie on the contour line, and the common end point P A denotes the point to be analyzed on the contour.
  • the lengths of a and b are arbitrary but fixed and can be determined depending on the noise level of the contour. Too large or too short a length will lead to an inaccurate determination of the orientation.
  • the contour line at p k is concave; r should be then are inverted so as not to point in the object's interior. If (t ⁇ , b) - 0, then the contour line at p k is straight and r can be determined as a vector perpendicular to a eg by (-a y , -a x ) ⁇ . In the latter case, the correct direction with respect to the object must be checked by other means. Place r equal to the length of the smallest cell diameter and check to see if the pixel determined by the vector is inside or outside the cell. The vector is inverted if it is inside the cell.
  • the optimal choice of the length of r depends on various parameters. If the maximum length of the fibers to be recognized is known, this value represents the optimum choice. In general, however, it should be taken into consideration that a length which is too short will result in a circular sector which does not completely cover the fiber space at the respective point p k ; As a result, important intensities in this area may not be transformed. In particular, for large opening angles ⁇ and p, this effect is amplified. In the case of smaller angles and a predominantly orthogonal orientation of the fibers, we suggest the Hausdorff distance [Hutrenlocher, DP, GA Klanderman and WJ Rucklidge: Comparing Images Using the Hausdorff Distance.
  • the image section under the circular sector is subjected to the linear Hough transformation with respect to p, so that a one-dimensional Hough space is created.
  • the space has a fixed height of 2 ⁇ or ⁇ + p.
  • the calculated one-dimensional parameter spaces are now combined into a superordinate space whose first dimension is the pixels of the cell contour.
  • the second dimension describes the opening angle.
  • the individual rooms are arranged in order of their calculation.
  • the advantage of the method according to the invention over the prior art is that not only a local model is used, but the calculated local parameter spaces are still correlated with each other. Considering two adjacent one-dimensional parameter spaces, a maximum (p "/ 0 m ) in the first space will most likely correspond to an actually existing line, even if the neighboring space also contains high values.
  • the method benefits from the fact that the potential starting points of the fiber lines can be determined a priori, so that the transformation can be restricted thereby.
  • the intensities of a given fiber will occur in part in the transformation in several consecutive parameter spaces. However, all intensities of the fiber occur only in the sector through whose origin the fiber passes. A maximum thus arises only in the corresponding parameter space.
  • the image of a cell 12 (FIG. 2a) with fibers 1.0 or cell traces is to be analyzed.
  • the image is subjected to a global threshold analysis to determine the space in which the fibers are to be located.
  • Two threshold values are determined such that one of the two delimits the background against fiber space intensities and the second the fiber intensities against the cell body. By applying the values one obtains in each case a binary image which contains the fiber space including the cell body or only the cell body. Subtracting both binary images provides the fiber space 14 as shown in FIG. 2a.
  • a circular sector is now calculated in order to locally limit the transformation to be subsequently performed.
  • Fig. 2c shows a selection of the detected sectors 18. Their opening angle indicates the slope range in which to search for fibers. After cutting the sectors with the fiber space, the transformation is performed. In Fig. 2 ⁇ the one-dimensional parameter spaces are shown after their combination. The coordinate system indicates horizontally the index k of the respective space 18; vertically the angle is shown. Black areas indicate that portion of the sectors 18 that are not transformed due to their limited opening angle.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

L'invention concerne un procédé de détection de structures géométriques dans des images, notamment dans des images d'échantillons chimiques et/ou biologiques, telles que des images de cellules, consistant à détecter une ligne de limitation d'un objet reproduit; à définir au moins un secteur (18) au sein de l'image, présentant une origine de secteur pk située sur la ligne de limitation (16); à transformer une section de l'image, définie par le secteur, en un espace de transformation au moyen d'une transformation affectant des signatures à des structures géométriques présentent dans ladite section de l'image, dans l'espace de transformation; à déterminer la présence d'une ou plusieurs signatures au sein de l'espace de transformation; et à transformer en retour des signatures dans le secteur, à partir de l'espace de transformation, pour la représentation de la structure géométrique.
EP05769803A 2004-07-10 2005-07-08 Procede de detection de structures geometriques dans des images Withdrawn EP1766579A2 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP05769803A EP1766579A2 (fr) 2004-07-10 2005-07-08 Procede de detection de structures geometriques dans des images

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
EP04016307A EP1615170A1 (fr) 2004-07-10 2004-07-10 Procédés de segmentation d'image, appliques en biologie cellulaire
PCT/EP2005/053275 WO2006005728A2 (fr) 2004-07-10 2005-07-08 Procede de detection de structures geometriques dans des images
EP05769803A EP1766579A2 (fr) 2004-07-10 2005-07-08 Procede de detection de structures geometriques dans des images

Publications (1)

Publication Number Publication Date
EP1766579A2 true EP1766579A2 (fr) 2007-03-28

Family

ID=34925710

Family Applications (2)

Application Number Title Priority Date Filing Date
EP04016307A Withdrawn EP1615170A1 (fr) 2004-07-10 2004-07-10 Procédés de segmentation d'image, appliques en biologie cellulaire
EP05769803A Withdrawn EP1766579A2 (fr) 2004-07-10 2005-07-08 Procede de detection de structures geometriques dans des images

Family Applications Before (1)

Application Number Title Priority Date Filing Date
EP04016307A Withdrawn EP1615170A1 (fr) 2004-07-10 2004-07-10 Procédés de segmentation d'image, appliques en biologie cellulaire

Country Status (3)

Country Link
US (1) US7903862B2 (fr)
EP (2) EP1615170A1 (fr)
WO (1) WO2006005728A2 (fr)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7817841B2 (en) 2005-11-12 2010-10-19 General Electric Company Time-lapse cell cycle analysis of unstained nuclei
US7508994B2 (en) * 2005-12-05 2009-03-24 Eastman Kodak Company Method for detecting streaks in digital images
CN102074011B (zh) * 2011-01-12 2012-05-30 河南理工大学 数字图像中任意三角形的检测方法
CN102096820B (zh) * 2011-01-12 2012-08-22 河南理工大学 数字图像中基于距离分布的正方形检测方法
CN106971400B (zh) * 2017-03-10 2020-11-10 中国航空工业集团公司洛阳电光设备研究所 一种图像分割线的修补方法及其装置
CN107872766A (zh) * 2017-10-20 2018-04-03 南京邮电大学 一种有向传感器网络节点感知区域相交方法

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6675040B1 (en) * 1991-01-28 2004-01-06 Sherwood Services Ag Optical object tracking system
DE10017551C2 (de) * 2000-04-08 2002-10-24 Carl Zeiss Vision Gmbh Verfahren zur zyklischen, interaktiven Bildanalyse sowie Computersystem und Computerprogramm zur Ausführung des Verfahrens
US7215802B2 (en) * 2004-03-04 2007-05-08 The Cleveland Clinic Foundation System and method for vascular border detection

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See references of WO2006005728A2 *

Also Published As

Publication number Publication date
US7903862B2 (en) 2011-03-08
US20070263917A1 (en) 2007-11-15
EP1615170A1 (fr) 2006-01-11
WO2006005728A3 (fr) 2006-03-23
WO2006005728A2 (fr) 2006-01-19

Similar Documents

Publication Publication Date Title
DE60036082T2 (de) Verfahren und vorrichtung zur wiederauffindung von texturbildern
DE19521346C2 (de) Bilduntersuchungs/-Erkennungsverfahren, darin verwendetes Verfahren zur Erzeugung von Referenzdaten und Vorrichtungen dafür
EP2187351A1 (fr) Procédé et dispositif destinés à la reconnaissance d'objets
EP1766579A2 (fr) Procede de detection de structures geometriques dans des images
EP3213294A1 (fr) Détermination d'un indice de qualité à partir d'un jeu de données d'une image volumique
DE602004009875T2 (de) Verfahren zur Bildverarbeitung für Profilbestimmung mittels strukturiertem Lichts
DE19928231C2 (de) Verfahren und Vorrichtung zur Segmentierung einer Punkteverteilung
DE102005025220B4 (de) Gerät, Verfahren und Programm zum Beseitigen von Poren
DE102006044595B4 (de) Bildverarbeitungsvorrichtung zur Segmentierung anhand von Konturpunkten
DE102017110339A1 (de) Computerimplementiertes Verfahren zur Vermessung eines Objekts aus einer digitalen Darstellung des Objekts
EP2622540A1 (fr) Procédé de classification de motifs dans des blocs de données d'image
DE102005049017A1 (de) Verfahren zur Segmentierung in einem n-dimensionalen Merkmalsraum und Verfahren zur Klassifikation auf Grundlage von geometrischen Eigenschaften segmentierter Objekte in einem n-dimensionalen Datenraum
DE10212277A1 (de) Verfahren zur Verbesserung von Fingerabdruck-Bildern
EP2096578A2 (fr) Procédé et dispositif de caractérisation de la formation de papier
EP3142068B1 (fr) Procede de detection tridimensionnelle d'objets
EP4185986A1 (fr) Procédé et système ou dispositif pour reconnaître un objet dans une image électronique
DE112004001486T5 (de) System und Verfahren zur Ortung von kompakten Objekten in Bildern
DE102018101014B3 (de) Verfahren zum Detektieren charakteristischer Merkmale eines Lichtmusters in einem mittels einer Fahrzeugkamera erfassten Abbild davon
EP3214602A1 (fr) Procédé de détection tridimensionnelle d'objets
EP2345004B1 (fr) Procédé et système de contrôle pour réaliser un contrôle optique du contour d'un objet
DE102012208025A1 (de) Verfahren und Vorrichtung zur rechnergestützten Segmentierung wenigstens eines einen Code mit wenigstens einer Codezeile enthaltenden Bildes
EP2887264B1 (fr) Détection de forme et de segment circulaire
EP3385910A1 (fr) Procédé d'identification des zones d'images correspondantes dans une séquence d'images
EP3685349A1 (fr) Procédé de détection d'une géométrie d'une zone partielle d'un objet
DE102005039424B4 (de) Verfahren und Vorrichtung zum Erkennen und zur Größenvermessung von Bereichen gleicher Textur in einem Digitalbild

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20070103

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LI LT LU LV MC NL PL PT RO SE SI SK TR

17Q First examination report despatched

Effective date: 20070522

DAX Request for extension of the european patent (deleted)
RIN1 Information on inventor provided before grant (corrected)

Inventor name: JETZEK, FRITZ

Inventor name: KIRSCH, ACHIM

RIN1 Information on inventor provided before grant (corrected)

Inventor name: JETZEK, FRITZ

Inventor name: KIRSCH, ACHIM

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20170110