US5933518A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Method and device for image processing for automatically detecting objects in digitized images
Assignee: US Philips Corp
•
Filing: 1996-04-22
•
Granted: 1999-08-03
•
Expires: 2016-04-22
Abstract
Abstract An image processing method is for automatically detecting objects of a predetermined type among diffuse objects in an original image, in the form of a two-dimensional matrix of pixels which have digitized intensity values. The method is carried out by a processing chain. A first transformed image (11A, 11B) is formed by a first smoothing (120A, 120B) of the original image (10A) to the resolution of the objects to be detected, in which local intensity optimums (LIST1A, LIST1B) are determined (130A, 140A; 130B, 140B). A second transformed image (13A, 13B) is formed via a second smoothing (220A, 220B) of the original image (10A) to the resolution of diffuse objects, in which homogeneous intensity regions are determined (230A, 230B), local intensity optimums (LIST1A, LIST1B) determined on the basis of the first transformed image are placed (240A, 240B) in the homogeneous regions determined on the basis of the second transformed image. One of these optimums per homogeneous region is selected (LIST2A, LIST2B), and objects among these optimums are additionally selected which constitute a list (LIST3A, LIST3B) of objects of a given type detected in the original image. The method is applicable to automatic recognition of suspect regions in radiologic images.
Inventors
- 👤 Eric Cohen-Solal
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-30 01:20
CPC Classifications
G; G06; G06T; G06T7/00; G06T7/0002; G06T7/0012; G06T2207/00; G06T2207/10; G06T2207/10116; G06T2207/20; G06T2207/20112; G06T2207/20152; G06T2207/30; G06T2207/30004; G06T2207/30068
Claims Analysis
Total Claims: 21
Claim 1
Independent Claim
184 words
A method of image processing for automatically detecting objects of a given type among diffuse objects in an image called original image in the form of a two-dimensional pixel matrix which has digitized intensity values, this method comprising, in the form of a processing chain, steps which include: a first transformation of the original image comprising a bandpass filtering to the resolution of the objects to be detected, to produce a first transformed image, determining local intensity maxima based on said first transformed image whose associated watershed regions have at least a predetermined size, a second transformation of the original image comprising a bandpass filtering to the resolution of diffuse objects, to produce a second transformed image, determining watershed regions in the second transformed image, placing the local intensity maxima determined on the basis of the first transformed image, in the watershed regions in the second transformed image, and selecting one of these maxima per watershed region, and forming a list of objects among the selected intensity maxima, which constitutes a list of the objects of a given type detected in the original image.
Claim 2
Independent Claim
184 words
A method of image processing for automatically detecting objects of a given type among diffuse objects in an image called original image in the form of a two-dimensional pixel matrix which has digitized intensity values, this method comprising, in the form of a processing chain, steps which include: a first transformation of the original image comprising a bandpass filtering to the resolution of the objects to be detected, to produce a first transformed image, determining local intensity maxima based on said first transformed image whose associated watershed regions have at least a predetermined size, a second transformation of the original image comprising a bandpass filtering to the resolution of diffuse objects, to produce a second transformed image, determining watershed regions in the second transformed image, placing the local intensity maxima determined on the basis of the first transformed image, in the watershed regions in the second transformed image, and selecting one of these maxima per watershed region, and forming a list of objects among the selected intensity maxima, which constitutes a list of the objects of a given type detected in the original image.
Claim 3
Dependent Claim
95 words
Method as claimed in claim 1, according to which the two-dimensional matrix is associated to orthogonal axes, and according to which the processing chain has a first form called first chain, in which: the first transformed image is obtained by a linear transformation including an isotropic band pass filtering with a core (σ) of the order of the dimension of the objects to be detected, the second transformed image is obtained by a linear transformation including an isotropic band pass filtering with a core (σ) of the order of the dimension of the diffuse objects.
Claim 4
Dependent Claim
132 words
Method as claimed in claim 1, according to which the two-dimensional matrix is associated to orthogonal axes, and according to which the processing chain has a second form called second chain in which: the first transformed image is obtained by a non-linear transformation including an isotropic band pass filtering with a core of the order of the dimension of the pixel, followed by an isotropic low pass filtering with a core of the order of the dimension of the objects to be detected, the second transformed image is obtained by a non-linear transformation including an isotropic band pass filtering with a core of the order of the dimension of the pixel, followed by an isotropic low pass filtering with a core (σ) of the order of the dimension of the diffuse objects.
Claim 5
Dependent Claim
124 words
Image processing method as claimed in claim 2, comprising, besides the first chain, a second form of the processing chain, called second chain, in which: the first transformed image is obtained by a non-linear filtering including an isotropic band pass filtering with a core of the order of the dimension of the pixel, followed by an isotropic low pass filtering with a core of the order of the dimension of the objects to be detected, the second transformed image is obtained by a non-linear filtering including an isotropic band pass filtering with a core of the order of the dimension of the pixel, followed by an isotropic low pass filtering with a core (σ) of the order of the dimension of the diffuse objects.
Claim 6
Dependent Claim
192 words
Method as claimed in claim 4, according to which in either chain: determining the local maxima includes the formation, based on the first transformed image, of a first intensity gradient image obtained by a band pass filtering oriented along the said axes of the two-dimensional matrix, followed by a "placing of fictitious dots" at a rate of one per pixel, forced to follow the slopes of the first intensity gradient image to determine the points of convergence of the slopes as local maxima, determining watershed regions includes the formation, based on the second transformed image, of a second intensity gradient image obtained by a band pass filtering oriented along the said axes of the two-dimensional matrix, followed by a "placing of fictitious dots" at a rate of one per pixel, forced to follow the slopes of the second intensity gradient image to determine the points of convergence of the slopes, the counting of the fictitious dots which converge towards each of these points to determine the range and number of pixels of the corresponding regions, and determining adjacent pixels that belong to different regions as boundaries of these regions called watershed regions.
Claim 7
Dependent Claim
70 words
Method as claimed in claim 5, according to which the detected lists of objects formed from the selection of the maxima in the first and second chains are called respectively, list of the first chain and list of the second chain, method further comprising a combining step for combining the list of the first chain and the list of the second chain forming a better list of the detected objects.
Claim 8
Dependent Claim
175 words
Method as claimed in claim 4, according to which, in the first chain, for producing the first transformed image of said first chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering with a Laplacian core of σ=7 pixels, for producing the second transformed image of said first chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=15 pixels, and according to which, in the second chain, for producing the first transformed image of said second chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=1 pixel, and the isotropic low-pass filtering is effected by a recursive two-dimensional filtering having a Gaussian core of σ=7 pixels, for producing the second transformed image of said second chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=1 pixel, and the isotropic low-pass filtering is effected by a recursive two-dimensional filtering having a Gaussian core of σ=15 pixels.
Claim 9
Dependent Claim
77 words
Method as claimed in claim 1, applied to an image acquired by a transmission through a region of a body and comprising a preliminary forming step for forming a boundary line between a zone of the image that corresponds to a part of the region of the body that has a substantially constant thickness, and an adjacent zone of the image that corresponds to another part of the region of the body that has a non-constant thickness.
Claim 10
Dependent Claim
44 words
Method as claimed in claim 8, according to which in the step for determining watershed regions, one potential watershed region is divided into two watershed regions when it is traversed by said boundary line, in which regions the local maxima are then placed individually.
Claim 11
Dependent Claim
94 words
Method as claimed in claim 8, comprising for the formation of the boundary line between the two zones: the formation of a histogram from the number of pixels as a function of the intensity levels of the original image, a smoothing of this histogram, the formation of histogram gradients, the selection of the largest of the histogram gradients for determining the slope maximums in the histogram, the selection of a first slope maximum found among the previously determined maximums in the histogram, by going from regions of little opacity to regions of growing opacity.
Claim 12
Dependent Claim
25 words
Method as claimed in claim 1, comprising in the processing chain, the generation of marks for letting detected objects stand out on the original image.
Claim 13
Dependent Claim
47 words
Auxiliary method for displaying suspect biological masses, comprising the formation of an image of a region inside a human body in the form of a two-dimensional pixel matrix which has digitized intensities, the processing of this image being effected in the steps as defined in claim 1.
Claim 14
Dependent Claim
71 words
Auxiliary device for displaying medical images, comprising: a system for producing data of a medical image of a part of the human body in the form of a two-dimensional pixel matrix which has digitized intensity values, a display system for displaying the image, and a processing system including a processor which has access to the image data and to the display system for implementing the method as claimed in claim 1.
Claim 15
Dependent Claim
175 words
Method as claimed in claim 5, according to which, in the first chain, for producing the first transformed image of said first chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering with a Laplacian core of σ=7 pixels, for producing the second transformed image of said first chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=15 pixels, and according to which, in the second chain, for producing the first transformed image of said second chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=1 pixel, and the isotropic low-pass filtering is effected by a recursive two-dimensional filtering having a Gaussian core of σ=7 pixels, for producing the second transformed image of said second chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=1 pixel, and the isotropic low-pass filtering is effected by a recursive two-dimensional filtering having a Gaussian core of σ=15 pixels.
Claim 16
Dependent Claim
175 words
Method as claimed in claim 6, according to which, in the first chain, for producing the first transformed image of said first chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering with a Laplacian core of σ=7 pixels, for producing the second transformed image of said first chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=15 pixels, and according to which, in the second chain, for producing the first transformed image of said second chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=1 pixel, and the isotropic low-pass filtering is effected by a recursive two-dimensional filtering having a Gaussian core of σ=7 pixels, for producing the second transformed image of said second chain, the isotropic band pass filtering is effected by a recursive two-dimensional filtering having a Laplacian core of σ=1 pixel, and the isotropic low-pass filtering is effected by a recursive two-dimensional filtering having a Gaussian core of σ=15 pixels.
Claim 17
Dependent Claim
45 words
Method as claimed in claim 5, according to which in either chain, for producing the first and second gradient images, the oriented band pass filtering is effected by a filtering having a low-pass Gaussian core followed by a filtering by an oriented first-order differential operator.
Claim 18
Dependent Claim
77 words
Method as claimed in claim 2, applied to an image acquired by a transmission through a region of a body and comprising a preliminary forming step for forming a boundary line between a zone of the image that corresponds to a part of the region of the body that has a substantially constant thickness, and an adjacent zone of the image that corresponds to another part of the region of the body that has a non-constant thickness.
Claim 19
Dependent Claim
187 words
The method of claim 1 wherein: determining the local maxima comprises the formation, based on the first transformed image, of a first intensity gradient image obtained by a band pass filtering oriented along the said axes of the two-dimensional matrix, followed by a "placing of fictitious dots" at a rate of one per pixel, forced to follow the slopes of the first intensity gradient image to determine the points of convergence of the slopes as local maxima, and determining watershed regions comprises the formation, based on the second transformed image, of a second intensity gradient image obtained by a band pass filtering oriented along the said axes of the two-dimensional matrix, followed by a "placing of fictitious dots" at a rate of one per pixel, forced to follow the slopes of the second intensity gradient image to determine the points of convergence of the slopes, the counting of the fictitious dots which converge towards each of these points to determine the range and number of pixels of the corresponding regions, and determining adjacent pixels that belong to different regions as boundaries of these regions called watershed regions.
Claim 20
Dependent Claim
40 words
The method of claim 18 wherein for producing the first and second intensity gradient images, the oriented band pass filtering is effected by a filtering having a low-pass Gaussian core followed by a filtering by an oriented first-order differential operator.
Claim 21
Dependent Claim
17 words
The method of claim 1 wherein the watershed regions of predetermined size include at least 600 pixels.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2002-08-02 | 2003-02-01 | 2003-08-03 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2006-08-02 | 2007-02-01 | 2007-08-03 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2010-08-02 | 2011-02-01 | 2011-08-03 | $7700.00 | Window Closed / Lapsed |