CN107146231A - Retinal image bleeding area segmentation method and device and computing equipment - Google Patents
Retinal image bleeding area segmentation method and device and computing equipment Download PDFInfo
- Publication number
- CN107146231A CN107146231A CN201710308401.7A CN201710308401A CN107146231A CN 107146231 A CN107146231 A CN 107146231A CN 201710308401 A CN201710308401 A CN 201710308401A CN 107146231 A CN107146231 A CN 107146231A
- Authority
- CN
- China
- Prior art keywords
- image
- pixel
- dark
- retinal images
- enhancing
- 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.)
- Granted
Links
- 230000004256 retinal image Effects 0.000 title claims abstract description 123
- 238000000034 method Methods 0.000 title claims abstract description 97
- 230000011218 segmentation Effects 0.000 title abstract description 11
- 230000000740 bleeding effect Effects 0.000 title abstract description 6
- 238000001914 filtration Methods 0.000 claims abstract description 47
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 41
- 230000002708 enhancing effect Effects 0.000 claims description 95
- 230000002008 hemorrhagic effect Effects 0.000 claims description 73
- 238000012545 processing Methods 0.000 claims description 35
- 238000009826 distribution Methods 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 12
- 230000002792 vascular Effects 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 8
- 238000002372 labelling Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 239000008280 blood Substances 0.000 claims description 4
- 210000004369 blood Anatomy 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 210000001525 retina Anatomy 0.000 abstract description 10
- 208000032843 Hemorrhage Diseases 0.000 abstract description 7
- 238000004891 communication Methods 0.000 description 16
- 238000003860 storage Methods 0.000 description 13
- 238000005516 engineering process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 208000037111 Retinal Hemorrhage Diseases 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000003902 lesion Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000004438 eyesight Effects 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000005728 strengthening Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 206010002329 Aneurysm Diseases 0.000 description 1
- 201000004569 Blindness Diseases 0.000 description 1
- 206010012689 Diabetic retinopathy Diseases 0.000 description 1
- 101000746134 Homo sapiens DNA endonuclease RBBP8 Proteins 0.000 description 1
- 101000969031 Homo sapiens Nuclear protein 1 Proteins 0.000 description 1
- 102100021133 Nuclear protein 1 Human genes 0.000 description 1
- 238000005267 amalgamation Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 206010012601 diabetes mellitus Diseases 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000007717 exclusion Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002207 retinal effect Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000004513 sizing Methods 0.000 description 1
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
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G06T5/70—
-
- G06T5/90—
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- 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
- G06T2207/20212—Image combination
- G06T2207/20224—Image subtraction
-
- 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
- G06T2207/30041—Eye; Retina; Ophthalmic
Abstract
The invention discloses a retinal image bleeding area segmentation method which is executed in computing equipment and comprises the following steps: obtaining a retina image to be segmented, and carrying out contrast enhancement on the image to obtain an enhanced image; filtering the enhanced image to extract a background image of the retina image; taking a difference value between the color value of each pixel in the enhanced image and the color value of the corresponding pixel in the background image to obtain a difference value image; obtaining a dark area image containing a dark area in the retina image according to the RGB color value of each pixel in the difference image, wherein the dark area comprises a blood vessel area, a bleeding area and a dark noise area; determining a blood vessel region from the enhanced image, and removing the region from the dark region image to obtain a blood vessel removed image; a dark noise region is determined from the enhanced image and removed from the de-angioed image, resulting in a hemorrhage region of the retinal image. The invention also discloses a corresponding retina image bleeding area segmentation device.
Description
Technical field
The present invention relates to the hemorrhagic areas dividing method of technical field of image processing, more particularly to a kind of retinal images.
Background technology
Diabetic retinopathy (referred to as sugared net) is a kind of ophthalmology disease being widely present in diabetes patient, its meeting pair
The eyesight of patient produces influence, and serious results even in blindness.Regular examination, find that PVR can be most as early as possible
Reduce patient's vision damage to big degree.Retinal hemorrhage lesion is the view caused by the rupture of intraretinal aneurysms
Film internal haemorrhage, it is one of sugar net early stage visible mark.Therefore, the blutpunkte in retinal images is accurately detected,
For realizing that automatic examination, the effective development assessed and suppress the state of an illness of sugar net are significant.
But, it is poor with background contrasts because the focus blur margin of retinal hemorrhage point is clear, with the gray scale of blood vessel excessively
It is close, it is in irregular shape and not of uniform size, and retinal images image quality uncertainty, cause to retinal hemorrhage area
The difficulty of domain automatic detection is very big, there is high false drop rate, loss height, computing complexity, the low problem for the treatment of effeciency.
Accordingly, it would be desirable to a kind of new not only accurate but also quick retinal images hemorrhagic areas dividing method.
The content of the invention
Therefore, the present invention provides a kind of retinal images hemorrhagic areas dividing method, device and computing device, to solve or
At least alleviate the problem of existing above.
According to an aspect of the present invention there is provided a kind of retinal images hemorrhagic areas dividing method, in computing device
Perform, this method includes：Retinal images to be split are obtained, and contrast enhancing is carried out to the image, the view is obtained
The enhancing image of film image；Processing is filtered to enhancing image, to extract the background image of the retinal images；It will increase
The RGB color value of the RGB color value and respective pixel in background image of each pixel takes difference in strong image, obtains error image；
Dark space area image is worth to according to the RGB color of each pixel in error image, retinal images have been remembered in the dark space area image acceptance of the bid
In dark areas, the dark areas include angiosomes, hemorrhagic areas and dark noise region；Area vasculosa is determined from enhancing image
Domain, and the region is removed from the area image of dark space, obtain blood-vessel image；And determine dark noise area from enhancing image
Domain, and the region is removed from blood-vessel image is removed, obtain the hemorrhagic areas of retinal images.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, to retinal images progress pair
The step of strengthening than degree includes：The color value of the RGB triple channels of each pixel in retinal images is normalized between 0~1
Number；For each Color Channel in RGB, the color value of each pixel in enhancing image is determined according to below equation：I1(x, y)=
α·I0(x,y)-β·I(x,y；δ)+γ；Wherein, I1(x, y) represents to strengthen the color value for the pixel that coordinate in image is (x, y),
I0(x, y) represents that coordinate is the color value of the pixel of (x, y), I (x, y in retinal images；δ) represent in retinal images
Coordinate for the pixel of (x, y) local mean value, wherein, local mean value is is δ gaussian filtering institute through window size and variance
Draw.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, enhancing image is filtered
The step of processing, background image to extract retinal images, includes：The multiple wave filters with different windows size of generation；
To tri- Color Channels of RGB of each pixel in enhancing image, multiple wave filters are respectively adopted processing is filtered to each passage, obtain
To multiple filter results of each passage；And multiple filter results of each passage are taken into the average color value as the passage, from
And obtain background image.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, Wiener filtering is filtered into, it is counted
Calculating formula is：
Wherein,The frequency-domain transform of the image extracted by Wiener filtering, G (u, v) is the current institute of Wiener filtering
The frequency-domain transform of image is handled, H (u, v) is degenrate function, and K is fixed constant.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, according to each picture in error image
The step of RGB color of element is worth to dark space area image includes：Obtain the color of the RGB triple channels of each pixel in error image
Value, and the color threshold of each passage is determined according to each passage color value got；And it is every in error image by contrasting
Each passage color values of RGB and the color threshold of respective channel of individual pixel, each passage color value of the pixel is marked be or
Mark is, so that error image is converted into dark space area image, the dark space area image is bianry image.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, blood is determined from enhancing image
Area under control domain, and include the step of the region is removed from the dark space area image：It is big using multiple windows under different variances
It is small that the enhancing image is repeatedly filtered, multiple filter results under each variance are respectively obtained, and this multiple filtering is tied
Fruit is averaged, and obtains the filtering average under the variance；Filtering average under each variance is merged, and to the image after merging
Enter row threshold division, obtain intermediate image, the intermediate image includes pseudo- angiosomes and angiosomes, and it is bianry image；
The pseudo- angiosomes in the image is determined by carrying out connected domain analysis to intermediate image；By pseudo- angiosomes from intermediate image
It is middle to remove, the distribution map of angiosomes is obtained, vascular distribution figure is designated as；And by the RGB color value of each pixel of dark space area image
Difference is taken with the RGB color value of respective pixel in vascular distribution figure, so that angiosomes be removed from the area image of dark space.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, by the intermediate image
Connected domain analysis is carried out the step of determining the pseudo- angiosomes in the image to include：Determine each connected domain in intermediate image；
The property value of each connected domain is calculated, wherein property value includes the area, girth, the minimum rectangle comprising the connected domain of the connected domain
Frame, and have with region in oval eccentricity, long axis length and the minor axis length of identical standard second-order moment around mean at least
It is a kind of；And judge whether meet the first predetermined condition between the property value of each connected domain, if so, being by the connected component labeling then
Pseudo- angiosomes.
Alternatively, according to the present invention retinal images hemorrhagic areas dividing method in, the first predetermined condition include with
Any one lower situation：The area of connected domain meets first threshold scope, and the area ratio of minimum rectangle frame and connected domain is more than
Second Threshold, and the ratio between long axis length and the minor axis length of connected domain are less than the 3rd threshold value；Or, the area of connected domain meets the
The area ratio of one threshold range, minimum rectangle frame and the connected domain is less than the ratio between the 4th threshold value, girth and is more than the 5th threshold value；
Or the area of connected domain is less than the 7th threshold value less than the 6th threshold value, eccentricity and the ratio between long axis length and minor axis length are less than
8th threshold value.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, determined from enhancing image dark
The step of noise region, includes：Enhancing image is gone into HSV color spaces by rgb color space；Judge each picture in enhancing image
Whether the HSV value of element meets the second predetermined condition；If so, the pixel then is labeled as into dark noise.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, determined from enhancing image dark
The step of noise region, also includes：Each pixel is calculated in enhancing image in the gradient magnitude of G passages, and each connected domain inside gradient
The average of amplitude, wherein connected domain are suitable to determine from the area image of dark space；And if the average of some connected domain inside gradient amplitude
Then it is dark noise region by the connected component labeling less than predetermined threshold.
Alternatively, in the retinal images hemorrhagic areas dividing method according to the present invention, α=β=4, γ=0.5, δ is
Arbitrary integer between 10~20；First threshold scope is [200,5000], and Second Threshold is 0.35, and the 3rd threshold value is 2.5, the
Four threshold values are 0.25, and the 5th threshold value is 0.95, and the 6th threshold value is 600, and the 7th threshold value is 0.97, and the 8th threshold value is 2.
According to an aspect of the present invention there is provided a kind of retinal images hemorrhagic areas segmenting device, reside at calculating and set
In standby, the device includes：Image pre-processing unit, suitable for obtaining retinal images to be split, and is contrasted to the image
Degree enhancing, obtains the enhancing image of the retinal images, and is filtered processing to enhancing image, to extract retina
The background image of image；Error image generation unit, suitable for that will strengthen in image in the RGB color value and background image of each pixel
The RGB color value of respective pixel takes difference, obtains error image；Dark areas determining unit, suitable for according to each picture in error image
The RGB color of element is worth to dark space area image, and the dark areas in retinal images, dark areas have been remembered in the dark space area image acceptance of the bid
Including angiosomes, hemorrhagic areas and dark noise region；Blood vessel removal unit, suitable for determining angiosomes from enhancing image,
And remove in the region from the area image of dark space, obtain blood-vessel image；And dark noise removal unit, suitable for from enhancing image
Middle determination dark noise region, and the region is removed from blood-vessel image is removed, obtain the hemorrhagic areas of retinal images.
According to an aspect of the present invention there is provided a kind of computing device, including：At least one processor；With the journey that is stored with
The memory of sequence instruction, described program instruction includes retinal images hemorrhagic areas as described above segmenting device；Wherein, handle
Device is configured as being suitable to being performed as described above according to the retinal images hemorrhagic areas segmenting device stored in the memory
Retinal images hemorrhagic areas dividing method.
It is described according to an aspect of the present invention there is provided a kind of computer-readable recording medium for the instruction that has program stored therein
Programmed instruction includes retinal images hemorrhagic areas as described above segmenting device；When storage in the computer-readable recording medium
Retinal images hemorrhagic areas segmenting device when being read by computing device, the computing device can be performed to be regarded as described above
Nethike embrane image hemorrhagic areas dividing method.
Original retinal images are carried out contrast enhancing by technique according to the invention scheme first, eliminate original view
The problems such as uneven illumination that film image is present, the contrast between enhancing hemorrhagic areas and background image makes follow-up hemorrhagic areas
Segmentation it is more accurate.Afterwards, by carrying out Wiener filtering to enhancing image, the Background of original retinal images has been extracted
Picture, and by taking difference to obtain error image enhancing image and background image.Afterwards, the dark areas threshold of the error image is obtained
Value, and then error image is converted into the dark space area image for including dark areas in retinal images, the dark areas according to the threshold value
Including angiosomes, hemorrhagic areas and dark noise region.
Then, pseudo- blood vessel is determined by carrying out gaussian filtering, Threshold segmentation and connected domain analysis to enhancing image successively
Region, and will the pseudo- angiosomes remove after obtain the distribution map of angiosomes, i.e. vascular distribution figure, and then by dark space area image
Take difference to remove angiosomes with vascular distribution figure, obtain blood-vessel image.Finally, color and gradient method is respectively adopted
To determine dark noise region, and the dark noise region is removed from blood-vessel image is removed, obtain final hemorrhagic areas.This method
Pass through repeatedly successively handling to relevant interference region so that the segmentation of hemorrhagic areas is more accurate, it is to avoid the mistake of hemorrhagic areas
Sentence, and the present invention also largely reduces the complexity of later image processing, accelerates calculating speed.
Brief description of the drawings
In order to realize above-mentioned and related purpose, some illustrative sides are described herein in conjunction with following description and accompanying drawing
Face, these aspects indicate the various modes of principles disclosed herein that can put into practice, and all aspects and its equivalent aspect
It is intended to fall under in the range of theme claimed.The following detailed description by being read in conjunction with the figure, the disclosure it is above-mentioned
And other purposes, feature and advantage will be apparent.Throughout the disclosure, identical reference generally refers to identical
Part or element.
Figure 1A shows segmenting system 100a in hemorrhagic areas according to an embodiment of the invention schematic diagram；
Figure 1B shows segmenting system 100b in hemorrhagic areas according to an embodiment of the invention schematic diagram；
Fig. 2 shows the structure chart of computing device 200 according to an embodiment of the invention；
Fig. 3 shows the structure chart of retinal images hemorrhagic areas according to an embodiment of the invention segmenting device 300；
Fig. 4 shows the flow chart of retinal images hemorrhagic areas according to an embodiment of the invention dividing method 400；
Fig. 5 A~Fig. 5 H show the effect for the embodiment split according to the retinal images hemorrhagic areas of the present invention
Figure.
Embodiment
The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although showing the disclosure in accompanying drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
Limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
Complete conveys to those skilled in the art.
Figure 1A shows segmenting system 100a in hemorrhagic areas according to an embodiment of the invention schematic diagram.Shown in Figure 1A
System 100a include retinal images collecting device 110 and computing device 200.It should be pointed out that the system 100a in Figure 1A is only
It is exemplary, in specific practice situation, any number of retinal images collecting device can be included in system 100a
110 and computing device 200, the present invention is to retinal images collecting device 110 and computing device 200 included in system 100a
Number be not limited.
Retinal images collecting device 110 for example can be the fundus camera of disposable type, and it is suitable to collection retina
Image；Computing device 200 can be the equipment such as PC, notebook computer, mobile phone, tablet personal computer, and it is adapted for carrying out at image
Reason task.In system 100a, distance of the retinal images collecting device 110 with computing device 200 spatially is closer, and two
Person can complete short-range communication in a wired or wireless manner, for example, retinal images collecting device 110 can pass through USB
Interface, RJ-45 interfaces, bnc interface etc. and computing device 200 set up wired connection, or by bluetooth, WiFi, ZigBee,
The agreements such as IEEE802.11x set up wireless connection with computing device 200, and the present invention is to retinal images collecting device 110 and meter
The connected mode calculated between equipment 200 is not limited.Retinal images hemorrhagic areas segmenting device is populated with computing device 200
300, device 300 can be or resident as a web application as an independent software installation in computing device 200
In the browser of computing device 200, or it is only one section of code in the memory of computing device 200, the present invention
Existence form of the device 300 in computing device 200 is not limited.When retinal images collecting device 110 collects view
After film image, retinal images are sent to computing device 200.Computing device 200 receives the retinal images, and by device
300 pairs of retinal images received are handled, and are partitioned into the hemorrhagic areas in retinal images.
Figure 1B shows segmenting system 100b in hemorrhagic areas according to an embodiment of the invention schematic diagram.Shown in Figure 1B
System 100b include retinal images collecting device 110, local client 120 and computing device 200.It should be pointed out that Figure 1B
In system 100b be only exemplary, in specific practice situation, any number of view can be included in system 100b
Film image collecting device 110, local client 120 and computing device 200, the present invention is to retina included in system 100b
The number of image capture device 110, local client 120 and computing device 200 is not limited.
Retinal images collecting device 110 for example can be the fundus camera of disposable type, and it is suitable to collection retina
Image；Local client 120 can be the equipment such as PC, notebook computer, mobile phone, tablet personal computer, and it is suitable to receive view
The retinal images that film image collecting device 110 is collected, and send it to computing device 200 via internet；Calculate
Equipment 200 can be implemented as server, and such as can be WEB server, apps server, it be adapted to provide for retina
The segmentation service of image hemorrhagic areas.In system 100b, retinal images collecting device 110 is with local client 120 in space
On distance it is closer, the two can complete short-range communication in a wired or wireless manner；Local client 120 is set with calculating
Standby 200 distance is distant, and the two can complete telecommunication via internet in a wired or wireless manner.Work as retina
Image capture device 110 is collected after retinal images, and retinal images are sent to local client 120.Then, local visitor
Family end 120 sends the retinal images received to computing device 200, the reception of computing device 200 retinal images, and by
300 pairs of retinal images received of device are handled, and are partitioned into the hemorrhagic areas in retinal images, and by segmentation result
It is back to local client 120.It should be understood that, although by retinal images collecting device 110 and native client in system 100b
End 120 is shown respectively as two equipment, still, it will be appreciated by those of skill in the art that in other examples, view
Film image collecting device 110 and local client 120 can be integrated into an equipment, and it is provided simultaneously with equipment described above
110 and local client 120 possess institute it is functional.
Fig. 2 shows the structure chart of computing device 200 according to an embodiment of the invention.In basic configuration 202, meter
Calculate equipment 200 and typically comprise system storage 206 and one or more central processing unit 204.Memory bus 208 can be with
For the communication between central processing unit 204 and system storage 206.Central processing unit 204 is the computing of computing device 200
Core and control core, its major function are interpretive machine instructions and handle data in various softwares.
Depending on desired configuration, central processing unit 204 can be any kind of processing, include but is not limited to：Micro- place
Manage device (μ P), microcontroller (μ C), digital information processor (DSP) or any combination of them.Central processing unit 204 can be with
Including such as cache of one or more rank of on-chip cache 210 and second level cache 212 etc, processing
Device core 214 and register 216.The processor core 214 of example can include arithmetic and logical unit (ALU), floating-point unit
(FPU), digital signal processing core (DSP core) or any combination of them.The Memory Controller 218 of example can be with
Central processing unit 204 is used together, or in some implementations, Memory Controller 218 can be the one of central processing unit 204
Individual interior section.
Depending on desired configuration, system storage 206 can be any type of memory, include but is not limited to：Easily
The property lost memory (RAM), nonvolatile memory (ROM, flash memory etc.) or any combination of them.System is stored
Device 206 can include operating system 220, one or more apply 222 and routine data 224.In some embodiments,
It may be arranged to be operated using routine data 224 on an operating system using 222.Using 222 bodies in the system memory
Now instructed for multi-segment program, for example, can be one section of JS code in executable program (.exe files) or webpage using 222.
Central processing unit 204 can perform these programmed instruction to realize using the function indicated by 222.In the present invention, apply
222 include retinal images hemorrhagic areas segmenting device 300.Retinal images hemorrhagic areas segmenting device 300 be one by
The instruction set of lines of code composition, it can indicate that central processing unit 204 performs the associative operation of image procossing, so as to realize
The hemorrhagic areas segmentation of retinal images.
Computing device 200 can also include contributing to from various interface equipments (for example, output equipment 242, Peripheral Interface
244 and communication equipment 246) to basic configuration 102 via the communication of bus/interface controller 230 interface bus 240.Example
Output equipment 242 include graphics processing unit 248 and audio treatment unit 250.They can be configured as contributing to via
One or more A/V port 252 is communicated with the various external equipments of such as display or loudspeaker etc.Outside example
If interface 244 can include serial interface controller 254 and parallel interface controller 256, they can be configured as contributing to
Via one or more I/O port 258 and such as input equipment (for example, keyboard, mouse, pen, voice-input device, touch
Input equipment) or the external equipment of other peripheral hardwares (such as printer, scanner) etc communicated.The communication of example is set
Standby 246 can include network controller 260, and it can be arranged to be easy to via one or more COM1 264 and one
The communication that other individual or multiple computing devices 262 pass through network communication link.
Network communication link can be an example of communication media.Communication media can be generally presented as in such as carrier wave
Or computer-readable instruction in the modulated data signal of other transmission mechanisms etc, data structure, program module, and can
With including any information delivery media." modulated data signal " can such signal, one in its data set or many
It is individual or it change can the mode of coding information in the signal carry out.As nonrestrictive example, communication media can be with
Include the wire medium of such as cable network or private line network etc, and it is such as sound, radio frequency (RF), microwave, infrared
(IR) the various wireless mediums or including other wireless mediums.Term computer readable storage medium used herein can be wrapped
Include both storage medium and communication media.According to a kind of embodiment, have program stored therein instruction, journey in computer-readable recording medium
Sequence instruction includes retinal images hemorrhagic areas segmenting device 300.When the device 300 stored in computer-readable recording medium
When being read by computing device 200, the central processing unit 204 of computing device 200 can perform corresponding retinal images bleeding area
Domain splitting method, to realize the segmentation of hemorrhagic areas in retinal images.
Fig. 3 shows the structure chart of retinal images hemorrhagic areas according to an embodiment of the invention segmenting device 300.
As shown in figure 3, device 300 include image pre-processing unit 320, error image generation unit 340, dark areas determining unit 360,
Blood vessel removal unit 380 and dark noise removal unit 390.
Image pre-processing unit 320 is suitable to obtain retinal images to be split, and carries out contrast enhancing to the image,
Obtain the enhancing image of the retinal images, and obtain the retinal images after being filtered processing to the enhancing image
Background image.
Wherein, retinal images to be split are the original retinal map that retinal images collecting device 110 is collected
Picture, as shown in Figure 5A, wherein left side is the retinal images with the circular angle of visual field, the retinal images are complete circle；
Right side is the retinal images for having the irregular angle of visual field, and the retinal images respectively lack a part in upper lower half circle.Of course for
The convenient follow-up alignment processing to image pixel, image pre-processing unit 320 first can be cut to the retinal images, and
Picture size after cutting is adjusted to by preliminary dimension using image interpolation method or other existing methods.Fig. 5 B are in Fig. 5 A
The figure of left and right two cut and size adjusting after figure, the image of the left and right subsequently occurred in 5C-5H two is base respectively
The design sketch obtained after being handled based on this two figures in Fig. 5 B, such as Fig. 5 C are that Fig. 5 B are carried out after contrast enhancing
Enhancing image, Fig. 5 D are the background images obtained after being filtered to Fig. 5 C.It should be appreciated that can also not cut actually
With sizing operation (omit Fig. 5 B), and every processing (such as contrast enhancing) is directly carried out on the basis of Fig. 5 A, it is simultaneously
The judged result of final hemorrhagic areas is not interfered with.
According to a kind of embodiment, image pre-processing unit 320 can be carried out according to following methods to original retinal images
Contrast strengthens：The color value of the RGB triple channels of each pixel in retinal images is normalized to the number between 0~1；For
The Color Channel of each in RGB, the color value of each pixel in enhancing image is determined according to below equation：
I1(x, y)=α I0(x,y)-β·I(x,y；δ)+γ
Wherein, I1(x, y) represents that coordinate is the color value of the pixel of (x, y), I in enhancing image0(x, y) is represented in view
Coordinate is the color value of the pixel of (x, y), I (x, y in film image；δ) expression coordinate in retinal images is the pixel of (x, y)
Local mean value, wherein, local mean value is is drawn through the gaussian filtering that window size and variance are δ.According to an implementation
Example, α=β=4, γ=0.5, δ is the arbitrary integer between 10~20, and these certain numerical value are exemplary illustration, actual behaviour
It can be arranged as required in work as other numerical value, the invention is not limited in this regard.In addition, it will be appreciated that being sat in enhancing image
It is designated as the pixel color value I of (x, y)1(x, y) is likely to occur the situation less than 0 or more than 1, handles for convenience, can will be small
Value in 0 is all set to 0, and the value that will be greater than 1 is all set to 1.
According to another embodiment, image pre-processing unit 320 can be filtered according to following methods to enhancing image
Processing：The multiple wave filters (wave filter of such as Wiener filtering) with different windows size of generation；To each pixel in enhancing image
Tri- Color Channels of RGB, this multiple wave filter is respectively adopted processing is filtered to each passage, obtain multiple filters of each passage
Ripple result；And multiple filter results of each passage are taken into the average color value as the passage, so as to obtain each pixel
Color value, these pixels are to constitute the background image.
Extracting the method for image background has a variety of, such as Wiener filtering, medium filtering, but in view of the meter of Wiener filtering
Calculate speed is not influenceed by filtering window size, and the filtering window of medium filtering is bigger, and calculating speed is slower, and efficiency is got over
It is low.Therefore, the filtering method selection Wiener filtering in the present invention, its calculation formula is as follows：
Wherein,The frequency-domain transform of the background image extracted by Wiener filtering, G (u, v) is enhancing image
Frequency-domain transform, H (u, v) is degenrate function, and K is fixed constant.H (u, v) and K can be arranged as required to as arbitrary value, for H
For (u, v), when calculating, it is necessary to be sampled to u and v, to generate a discrete filter, the big I of filtering window
To be arranged as required to as arbitrary value.Drawn according to above formulaAfterwards, it is rightCarry out Fourier inversion, you can
Go out the background image of spatial domain.
Here it is possible to select these three filtering window sizes of 50*50,100*100 and 500*500, and filtered with these three
Ripple device window is filtered to each passage of the RGB of each pixel in enhancing image respectively.Here, in RGB triple channels
Each passage, will be respectively filtered with three kinds of wave filters, that is, be intended to be calculated with above-mentioned Wiener filtering formula, then will
The average value of the filter result of three wave filters as the passage color value.Method obtains each passage of each pixel according to this
Value, you can obtain final background image.
Strengthen after image and background image generation, error image generation unit 340 will strengthen the RGB face of each pixel in image
The RGB color value of colour and respective pixel in background image takes difference, obtains error image.Specifically, image will can be strengthened
In the RGB color value of each pixel subtract the RGB color value of respective pixel in background image.
Then, dark areas determining unit 360 is worth to dark space area image according to the RGB color of each pixel in error image,
The dark areas in original retinal images is remembered in this dark space area image acceptance of the bid, and the dark areas includes angiosomes, bleeding area
Domain and dark noise region.
Specifically, dark areas determining unit 360 is suitable to the color value for obtaining the RGB triple channels of each pixel in error image,
And the color threshold of the passage is determined according to each passage color value got；And by contrasting each picture in error image
Each passage color value of element and the color threshold of respective channel, each passage color value of the pixel is marked and is or marks
For 1, so that error image is converted into dark space area image, the dark space area image is bianry image.
It is possible to further (can using the color threshold of color average as the respective channel of each passage of all pixels
To be considered clear zone threshold value), will if the value of tri- passages of RGB of some pixel is all higher than the color threshold of respective channel
The rgb value of the pixel is set to 1；Otherwise, it is set to 0, so as to obtain a secondary binary map (i.e. artwork master), at this moment dark
Region be in the bianry image rgb value be set to 1 pixel set.It is of course also possible to take another assignment mode：Will
The former rgb value is set to 0, and the rgb value of the latter is set into 1, dark areas at this moment be in bianry image rgb value be 0
Pixel set.Fig. 5 E show the clear zone area image obtained according to the first assignment mode, i.e. bianry image, therein white
Color region is dark areas (rgb value is 1), and it corresponds to region dark in former retinal images.
Then, blood vessel removal unit 380 determines angiosomes from enhancing image, and by the region from the area image of dark space
Remove, obtain blood-vessel image.
Specifically, blood vessel removal unit 380 can determine angiosomes according to following methods from enhancing image：Using not
Enhancing image is repeatedly filtered with multiple window sizes under variance, multiple filter results under each variance are respectively obtained,
And be averaged this multiple filter result, obtain the filtering average under the variance.Afterwards, the filtering average under each variance is carried out
Merge, and row threshold division is entered to the image after merging, obtain an intermediate image, the intermediate image is bianry image, including
Pseudo- angiosomes and angiosomes.Finally, the pseudo- blood in the image is determined by carrying out connected domain analysis to the intermediate image
Area under control domain, and the pseudo- angiosomes is removed from intermediate image, the distribution map of angiosomes is obtained, vascular distribution figure is designated as.
Specifically, multiple window sizes of the blood vessel removal unit 380 under using different variances carry out many to enhancing image
During secondary filtering, the gaussian filtering of multiple window sizes can be used, this is mainly considered in retinal images
Local direction, the Curvature varying of medium vessels are smaller, and the grey scale change of cross section is approximately Gaussian curve.According to a kind of embodiment,
19 kinds of window sizes can be chosen under two variance δ, each variance using choosing herein, such as：
1) δ 1=5, and this 19 kinds of window sizes of 2*2~20*20 are chosen respectively to enhancing image progress gaussian filtering, obtain
19 filter results；
2) δ 2=1.8, and this 19 kinds of window sizes of 2*2~20*20 are chosen respectively to enhancing image progress gaussian filtering, obtain
To 19 filter results.
Afterwards, the average value of 19 kinds of filter results under the two variances is tried to achieve respectively, and try to achieve two filtering are equal
Value is merged, you can the figure after obtaining corresponding to amalgamation result.Have many to the algorithm that the image after merging enters row threshold division
Kind, such as OTSU Otsu algorithms, maximum entropy method (MEM), iterative method etc. can also refer to the generation method of dark space area image, i.e., first ask
The color threshold of the color value of each passage and each passage in the image after the merging is obtained, and by contrasting the color of each passage
The image after merging is converted to intermediate image by the color threshold of value and the passage, and intermediate image is marked angiosomes
With the bianry image of pseudo- angiosomes.
Further, blood vessel removal unit 380 can determine the pseudo- angiosomes in intermediate image according to following methods：
First determine algorithm using existing any connected domain to determine each connected domain in the intermediate image, then calculate each connected domain
Property value, wherein property value include area, girth, the minimum rectangle frame comprising the connected domain of the connected domain, and and region
At least one of oval eccentricity, long axis length and minor axis length with identical standard second-order moment around mean.Finally, judge
Whether first predetermined condition is met between the property value of each connected domain, if so, being then pseudo- angiosomes by the connected component labeling.Its
In, the first predetermined condition can be any one following condition：
1) area of connected domain meets first threshold scope, and the area ratio of minimum rectangle frame and connected domain is more than the second threshold
Value, and the ratio between long axis length and the minor axis length of connected domain are less than the 3rd threshold value；
2) area of connected domain meets first threshold scope, and the area ratio of minimum rectangle frame and connected domain is less than the 4th threshold
The ratio between value, girth are more than the 5th threshold value；Or
3) area of connected domain be less than the 6th threshold value, eccentricity be less than the 7th threshold value and long axis length and minor axis length it
Than less than the 8th threshold value.
Here, first threshold may range from [200,5000], and Second Threshold can be 0.35, and the 3rd threshold value can be
2.5, the 4th threshold value can be 0.25, and the 5th threshold value can be 0.95, and the 6th threshold value can be 600, and the 7th threshold value can be
0.97, the 8th threshold value can be 2.Of course, it is possible to be set to other numerical value as needed, the present invention is to these concrete numerical values
It is not construed as limiting.
It is determined that after pseudo- angiosomes, blood vessel removal unit 380 removes in the region from intermediate image, to obtain area vasculosa
The distribution map in domain, is designated as vascular distribution figure.It is the bianry image of a width black and white to consider intermediate image, and above in dark space area image
Generating process in the rgb value of the pixel of dark areas is set to 1, therefore can use pseudo- area vasculosa in intermediate image here
The rgb value of pixel in domain is set to 0 method to remove pseudo- angiosomes, obtained vascular distribution figure as illustrated in figure 5f, its
In white portion be angiosomes.In addition it is also possible to the coordinate value of each pixel in pseudo- angiosomes is recorded, and in centre
The rgb value of the pixel at respective coordinates is disposed as 0 in image.If of course it is to be understood that dark areas figure above
As the pixel color value of dark areas is set into 0 in generating process, then here just can be by the pixel color of pseudo- angiosomes
Value is set to 1.
Determine after angiosomes, blood vessel removal unit 380 can be by the RGB color value and blood vessel of each pixel of dark space area image
The RGB color value of respective pixel takes difference in distribution map, so that angiosomes be removed from the area image of dark space, what is obtained removes blood
Pipe image is as depicted in fig. 5g.Here, it can be that the RGB color value of each pixel of dark space area image is subtracted into blood vessel point to take difference arithmetic
The RGB color value of respective pixel in Butut, naturally it is also possible to carry out taking difference after some weight calculations again, the present invention is to taking difference
The specific algorithm of process is not construed as limiting.
The hemorrhagic areas and dark noise region for going the white portion in blood-vessel image to correspond in former retinal images, are removed
Dark noise region is that can obtain hemorrhagic areas.Therefore, dark noise removal unit 390 determines dark noise region from enhancing image,
And remove in the region from blood-vessel image is removed, so as to obtain the hemorrhagic areas of retinal images.
According to one embodiment, dark noise removal unit 390 can determine dark noise region according to the method for color.Tool
Body, HSV color spaces are gone to by enhancing image by rgb color space, and judge the HSV value of each pixel whether to meet second pre-
Fixed condition, if so, the pixel then is labeled as into dark noise.Wherein, the second predetermined condition can be any one following situation：H
Value is outside first interval scope, or S values are outside second interval scope, or V values are outside 3rd interval scope.Wherein, the firstth area
Between scope be [0.45,1], second interval scope be [0.15,0.75], 3rd interval scope be [0.45,0.75].Here, when
When HSV value is all in correspondence interval range, it is believed that the pixel is blutpunkte.It should be pointed out that this, which is in, judges that HSV value is
During the second predetermined condition of no satisfaction, the HSV value used is the HSV value after normalization, i.e. HSV value is normalized to 0 first~
Number between 1, then judge whether the HSV value after normalization meets the second predetermined condition.In addition, it will be appreciated that these numerical value are only
It is exemplary illustration, can be arranged as required in practical operation as other numerical value, the invention is not limited in this regard.
Determined according to color approach after dark noise, it is possible to remove the dark noise in blood-vessel image is removed.Here may be used
With the minimizing technology with reference to pseudo- angiosomes, equally this is removed using the pixel color value of dark noise is set into 0 or 1 method
Dark noise.In addition it is also possible to the coordinate value for the pixel for being marked as dark noise be recorded, then to the seat in blood-vessel image is removed
The rgb value of pixel at mark is configured, and here is omitted.
According to another embodiment, dark noise removal unit 390 can also determine dark noise area according to the method for gradient
Domain.Specifically, it is possible to use the characteristics of color contrast of blutpunkte and background area is higher in G passages, enhancing image is calculated
In each pixel G passages gradient magnitude, and each connected domain inside gradient amplitude average, wherein connected domain can be from dark areas
Determined in image.It is dark noise by the connected component labeling if the average of some connected domain inside gradient amplitude is less than predetermined threshold
Region；It is on the contrary then be believed that the connected domain be hemorrhagic areas.Wherein, predetermined threshold is by all connected domains in the enhancing image
The average of the gradient magnitude of every other pixel after exclusion.Determined behind dark noise region, can equally joined according to gradient method
The minimizing technology for examining pseudo- angiosomes removes in the dark noise region marked from blood-vessel image is removed, and here is omitted.
In practical operation, the method for any selection color or gradient can be selected to determine dark noise region.But in order to
More preferable dark noise removal effect is realized, the dark noise region that can also be determined to both modes takes intersection.When using group
During the method for conjunction, both the above method can be implemented with random order, and the present invention is not limited to the sequencing of each method.It
Afterwards, dark noise region both approaches marked is removed from blood-vessel image is removed, so as to obtain final hemorrhagic areas.
In the combined method of two methods used herein, the step of determining dark noise includes two steps, and each step is in back
On the basis of determine to be determined as in dark noise, back that the region of dark noise no longer carries out repeating judgement in latter step, this
Sample, can either accurately determine dark noise region, avoid omitting, meanwhile, unnecessary calculating is decreased, so as to accelerate meter
Calculate speed.The hemorrhagic areas obtained by the combined treatment of both modes is as shown in the black spots dotted region in Fig. 5 H.
Fig. 4 shows the flow chart of retinal images hemorrhagic areas according to an embodiment of the invention dividing method 400.
Method 400 is suitable to perform in the retinal images hemorrhagic areas segmenting device 300 shown in earlier figures 3.As shown in figure 4, method
400 start from step S420.
In the step s 420, retinal images to be split are obtained, and contrast enhancing is carried out to the image, view is obtained
The enhancing image of film image；And in step S440, processing is filtered to enhancing image, to extract the back of the body of retinal images
Scape image.The detailed process of two steps may be referred to the foregoing description to image pre-processing unit 320, no longer go to live in the household of one's in-laws on getting married herein
State.
Then, in step S460, the RGB color value of each pixel and respective pixel in background image in image will be strengthened
RGB color value takes difference, obtains error image.The detailed process of the step may be referred to foregoing to error image generation unit
340 description, here is omitted.
In step S470, dark space area image is worth to according to the RGB color of each pixel in error image, the dark areas figure
The dark areas in retinal images is marked as in, wherein dark areas includes angiosomes, hemorrhagic areas and dark noise region.
The detailed process of the step may be referred to the foregoing description to dark areas determining unit 360, and here is omitted.
Then, in step S480, angiosomes is determined from enhancing image, and by the region from the area image of dark space
Remove, obtain blood-vessel image.The detailed process of the step may be referred to the foregoing description to blood vessel removal unit 380, herein not
Repeat again.
Then, in step S490, dark noise region is determined from enhancing image, and by the region from blood-vessel image is removed
Remove the hemorrhagic areas for obtaining retinal images.The detailed process of the step may be referred to foregoing to dark noise removal unit 390
Description, here is omitted.
The following is the embodiment of the retinal images hemorrhagic areas segmentation of the present invention：
1) original retinal images are obtained, as shown in Figure 5A；
2) original retinal images are cut and size adjusting, obtains Fig. 5 B；
3) contrast enhancing is carried out to Fig. 5 B, obtains enhancing image-Fig. 5 C；
4) Wiener filtering is carried out to Fig. 5 C, extracts background image-Fig. 5 D；
5) the rgb pixel value for strengthening image (Fig. 5 C) is subtracted to the rgb value of the respective pixel of background image (Fig. 5 D), obtained
Dark space area image, as shown in fig. 5e；
6) multiple gaussian filtering is carried out to Fig. 5 C, multiple filter result merged, and the figure after merging is converted to
The intermediate image represented by two-value, connected domain analysis is carried out to bianry image, it is determined that pseudo- angiosomes, and then by the pseudo- area vasculosa
Domain is removed from intermediate image, obtains vascular distribution figure-Fig. 5 F；
7) dark space area image (Fig. 5 E) is subtracted into vascular distribution figure (Fig. 5 F), obtains blood-vessel image, as depicted in fig. 5g；
8) dark noise region therein is determined according to Fig. 5 C color and gradient respectively, and by the dark noise region from Fig. 5 G
It is middle to remove, obtain final hemorrhagic areas as illustrated in fig. 5h.
Technique according to the invention scheme, enhancing image has been obtained by the way that retinal images are carried out with contrast enhancing, right
Enhancing image is filtered processing and has obtained background image；Error image has been obtained based on enhancing image and background image；From difference
The dark areas for including angiosomes, hemorrhagic areas and dark noise region has been isolated in value image.Afterwards, it is fixed from enhancing image
Position angiosomes and dark noise region；Finally angiosomes and dark noise region are removed from the area image of dark space, obtained most
Whole hemorrhagic areas.This method can quick, comprehensive lesion candidate region and interference region, and in lesion candidate regions
Domain removes interference region, so as to highly precisely be positioned to hemorrhagic areas, it is therefore prevented that the erroneous judgement to result.Moreover, this hair
The bright analysis detection speed for also improving traditional eye fundus image, greatly reduces manpower and materials during data processing, Neng Gouguang
The general automation application suitable for extensive eye fundus image.
In addition, the present invention is cut by the standardization to eye fundus image can reduce the difference between different eye fundus images,
And the enhancing processing of eye fundus image can reduce the difference that same eye fundus image is introduced due to uneven illumination, these are all from items
The processing accuracy of eye fundus image is improved in details, so as to further increase the accuracy of hemorrhagic areas judged result.
A8, the method as described in A7, wherein, first predetermined condition includes any one following situation：The connection
The area in domain meets first threshold scope, and the area ratio of the minimum rectangle frame and the connected domain is more than Second Threshold, and
The ratio between long axis length and minor axis length of the connected domain are less than the 3rd threshold value；The area of the connected domain meets first threshold model
Enclose, the area ratio of the minimum rectangle frame and the connected domain is less than the ratio between the 4th threshold value, girth and is more than the 5th threshold value；Or
The area of the connected domain is less than the 7th threshold value less than the 6th threshold value, eccentricity and the ratio between long axis length and minor axis length are less than
8th threshold value.
A9, the method as described in A1, wherein, the step of determining the dark noise region from the enhancing image includes：
The enhancing image is gone into HSV color spaces by rgb color space；And judge whether the HSV value of each pixel meets second
Predetermined condition, if so, the pixel then is labeled as into dark noise.
A10, the method as described in A1 or A9, wherein, the step of determining the dark noise region from the enhancing image
Also include：The gradient magnitude of each pixel in the enhancing image in G passages is calculated, and each connected domain inside gradient amplitude is equal
Value, the connected domain is suitable to determine from the dark space area image；If the average of some connected domain inside gradient amplitude is less than predetermined
Threshold value, then be dark noise region by the connected component labeling.
A11, the method as any one of A1-A10, wherein, α=β=4, γ=0.5, δ is times between 10~20
Meaning integer；The first threshold scope is [200,5000], and the Second Threshold is 0.35, and the 3rd threshold value is 2.5, described
4th threshold value is 0.25, and the 5th threshold value is 0.95, and the 6th threshold value is 600, and the 7th threshold value is 0.97, described the
Eight threshold values are 2.
B13, the device as described in B12, wherein, described image pretreatment unit is suitable to according to following methods to the view
Film image carries out contrast enhancing：The color value of the RGB triple channels of each pixel in the retinal images is normalized to 0~1
Between number；For each Color Channel in RGB, the color value of each pixel in enhancing image is determined according to below equation：I1
(x, y)=α I0(x,y)-β·I(x,y；δ)+γ wherein, I1(x, y) represents the pixel that the enhancing image coordinate is (x, y)
Color value, I0(x, y) represents that coordinate is the color value of the pixel of (x, y), I (x, y in the retinal images；δ) represent
In the retinal images coordinate for the pixel of (x, y) local mean value, wherein, the local mean value be through window size and
Variance is that δ gaussian filtering is drawn.
B14, the device as described in B12, wherein, described image pretreatment unit is suitable to according to following methods to the enhancing
Image is filtered processing：The multiple wave filters with different windows size of generation；To the RGB of each pixel in the enhancing image
Three Color Channels, are respectively adopted the multiple wave filter and are filtered processing to each passage, obtain multiple filtering of each passage
As a result；And multiple filter results of each passage are taken into the average color value as the passage, so as to obtain the background image.
B15, the device as described in B12 or B14, wherein described be filtered into Wiener filtering, its formula is：
Wherein,The frequency-domain transform of the image extracted by Wiener filtering, G (u, v) is the current institute of Wiener filtering
The frequency-domain transform of image is handled, H (u, v) is degenrate function, and K is fixed constant.
B16, the device as described in B12, wherein, the dark areas determining unit is suitable to obtain dark areas according to following methods
Image：Obtain the color value of the RGB triple channels of each pixel in the error image, and according to each passage color value got come
Determine the color threshold of the passage；And led to by contrasting each passage color value of each pixel in the error image with corresponding
The color threshold in road, each passage color value of the pixel is marked be or mark be so that the error image convert
For dark space area image, the dark space area image is bianry image.
B17, the device as described in B12, wherein, the blood vessel removal unit is suitable to from the enhancing be schemed according to following methods
The angiosomes is determined as in, and the region is removed from the dark space area image：Using multiple windows under different variances
Mouth size is repeatedly filtered to the enhancing image, respectively obtains multiple filter results under each variance, and will be the multiple
Filter result is averaged, and obtains the filtering average under the variance；Filtering average under each variance is merged, and to merging after
Image enter row threshold division, obtain intermediate image, the intermediate image includes pseudo- angiosomes and angiosomes, and it is two-value
Image；The pseudo- angiosomes in the image is determined by carrying out connected domain analysis to the intermediate image；By the pseudo- blood vessel
Region is removed from the intermediate image, obtains the distribution map of the angiosomes, is designated as vascular distribution figure；And by by institute
The RGB color value for stating the RGB color value and respective pixel in the vascular distribution figure of each pixel of dark space area image takes difference, will
The angiosomes is removed from the dark space area image.
B18, the device as described in B17, wherein, the blood vessel removal unit is further adapted for determining puppet according to following methods
Angiosomes：Determine each connected domain in the intermediate image；The property value of each connected domain is calculated, the property value includes the company
Area, girth, the minimum rectangle frame comprising the connected domain in logical domain, and there is the ellipse of identical standard second-order moment around mean with region
At least one of round eccentricity, long axis length and minor axis length；And judge whether expire between the property value of each connected domain
The first predetermined condition of foot, if so, being then pseudo- angiosomes by the connected component labeling.
B19, the device as described in B18, wherein, first predetermined condition includes any one following situation：The company
The area in logical domain meets first threshold scope, and the minimum rectangle frame and the area ratio of the connected domain are more than Second Threshold,
And the ratio between the long axis length of the connected domain and minor axis length are less than the 3rd threshold value；The area of the connected domain meets first threshold
The area ratio of scope, the minimum rectangle frame and the connected domain is less than the ratio between the 4th threshold value, girth and is more than the 5th threshold value；Or
The area of connected domain described in person is less than the 6th threshold value, eccentricity less than the 7th threshold value and the ratio between long axis length and minor axis length are small
In the 8th threshold value.
B20, the device as described in B12, wherein, the dark noise removal unit is suitable to according to following methods from the enhancing
The dark noise region is determined in image：The enhancing image is gone into HSV color spaces by rgb color space；And judge
Whether the HSV value of each pixel meets the second predetermined condition, if so, the pixel then is labeled as into dark noise.
B21, the device as described in B12 or B20, wherein, the dark noise removal unit is further adapted for according to lower section
Method determines the dark noise region from the enhancing image：Calculate gradient width of each pixel in G passages in the enhancing image
Value, and each connected domain inside gradient amplitude average, the connected domain be suitable to from the dark space area image determine；If some connects
The average of logical domain inside gradient amplitude is less than predetermined threshold, then is dark noise region by the connected component labeling.
B22, the device as any one of B12-B21, wherein, α=β=4, γ=0.5, δ be 10~20 between
Arbitrary integer；The first threshold scope is [200,5000], and the Second Threshold is 0.35, and the 3rd threshold value is 2.5, institute
It is 0.25 to state the 4th threshold value, and the 5th threshold value is 0.95, and the 6th threshold value is 600, and the 7th threshold value is 0.97, described
8th threshold value is 2.
Various technologies described herein can combine hardware or software, or combinations thereof is realized together.So as to the present invention
Method and apparatus, or the process and apparatus of the present invention some aspects or part can take embedded tangible media, such as it is soft
The form of program code (instructing) in disk, CD-ROM, hard disk drive or other any machine readable storage mediums,
Wherein when program is loaded into the machine of such as computer etc, and when being performed by the machine, the machine becomes to put into practice this hair
Bright equipment.
In the case where program code is performed on programmable computers, computing device generally comprises processor, processor
Readable storage medium (including volatibility and nonvolatile memory and/or memory element), at least one input unit, and extremely
A few output device.Wherein, memory is arranged to store program codes；Processor is arranged to according to the memory
Instruction in the described program code of middle storage, performs the retinal images hemorrhagic areas dividing method of the present invention.
By way of example and not limitation, computer-readable medium includes computer-readable storage medium and communication media.Calculate
Machine computer-readable recording medium includes computer-readable storage medium and communication media.Computer-readable storage medium storage such as computer-readable instruction,
The information such as data structure, program module or other data.Communication media is general modulated with carrier wave or other transmission mechanisms etc.
Data-signal processed passes to embody computer-readable instruction, data structure, program module or other data including any information
Pass medium.Any combination above is also included within the scope of computer-readable medium.
This place provide specification in, algorithm and display not with any certain computer, virtual system or other
Equipment is inherently related.Various general-purpose systems can also be used together with the example of the present invention.As described above, construct this kind of
Structure required by system is obvious.In addition, the present invention is not also directed to any certain programmed language.It should be understood that can
To realize the content of invention described herein using various programming languages, and the description done above to language-specific be for
Disclose the preferred forms of the present invention.
In the specification that this place is provided, numerous specific details are set forth.It is to be appreciated, however, that the implementation of the present invention
Example can be put into practice in the case of these no details.In some instances, known method, knot is not been shown in detail
Structure and technology, so as not to obscure the understanding of this description.
Similarly, it will be appreciated that in order to simplify the disclosure and help to understand one or more of each inventive aspect, exist
Above in the description of the exemplary embodiment of the present invention, each feature of the invention is grouped together into single implementation sometimes
In example, figure or descriptions thereof.However, the method for the disclosure should be construed to reflect following intention：It is i.e. required to protect
The application claims of shield are than the feature more features that is expressly recited in each claim.More precisely, as following
As claims reflect, inventive aspect is all features less than single embodiment disclosed above.Therefore, abide by
Thus the claims for following embodiment are expressly incorporated in the embodiment, wherein each claim is in itself
It is used as the separate embodiments of the present invention.
Those skilled in the art should be understood the module or unit or group of the equipment in example disclosed herein
Part can be arranged in equipment as depicted in this embodiment, or alternatively can be positioned at and the equipment in the example
In different one or more equipment.Module in aforementioned exemplary can be combined as a module or be segmented into addition multiple
Submodule.
Those skilled in the art, which are appreciated that, to be carried out adaptively to the module in the equipment in embodiment
Change and they are arranged in one or more equipment different from the embodiment.Can be the module or list in embodiment
Member or component be combined into a module or unit or component, and can be divided into addition multiple submodule or subelement or
Sub-component.In addition at least some in such feature and/or process or unit exclude each other, it can use any
Combination is disclosed to all features disclosed in this specification (including adjoint claim, summary and accompanying drawing) and so to appoint
Where all processes or unit of method or equipment are combined.Unless expressly stated otherwise, this specification (including adjoint power
Profit is required, summary and accompanying drawing) disclosed in each feature can or similar purpose identical, equivalent by offer alternative features come generation
Replace.
Although in addition, it will be appreciated by those of skill in the art that some embodiments described herein include other embodiments
In included some features rather than further feature, but the combination of the feature of be the same as Example does not mean in of the invention
Within the scope of and form different embodiments.For example, in the following claims, times of embodiment claimed
One of meaning arbitrarily combination can be used.
In addition, be described as herein can be by the processor of computer system or by performing for some in the embodiment
Method or the combination of method element that other devices of the function are implemented.Therefore, with for implementing methods described or method
The processor of the necessary instruction of element forms the device for implementing this method or method element.In addition, device embodiment
Element described in this is the example of following device：The device is used to implement as in order to performed by implementing the element of the purpose of the invention
Function.
As used in this, unless specifically stated so, come using ordinal number " first ", " second ", " the 3rd " etc.
Description plain objects are merely representative of the different instances for being related to similar object, and are not intended to imply that the object being so described must
Must have the time it is upper, spatially, in terms of sequence or given order in any other manner.
Although describing the present invention according to the embodiment of limited quantity, above description, the art are benefited from
It is interior it is clear for the skilled person that in the scope of the present invention thus described, it can be envisaged that other embodiments.Additionally, it should be noted that
The language that is used in this specification primarily to readable and teaching purpose and select, rather than in order to explain or limit
Determine subject of the present invention and select.Therefore, in the case of without departing from the scope and spirit of the appended claims, for this
Many modifications and changes will be apparent from for the those of ordinary skill of technical field.For the scope of the present invention, to this
The done disclosure of invention is illustrative and be not restrictive, and it is intended that the scope of the present invention be defined by the claims appended hereto.
Claims (10)
1. a kind of retinal images hemorrhagic areas dividing method, is performed in computing device, this method includes：
Retinal images to be split are obtained, and contrast enhancing is carried out to the image, the enhancing of the retinal images is obtained
Image；
Processing is filtered to the enhancing image, to extract the background image of the retinal images；
The RGB color value of the RGB color value and respective pixel in the background image of each pixel in the enhancing image is taken into difference
Value, obtains error image；
Dark space area image is worth to according to the RGB color of each pixel in the error image, the dark space area image acceptance of the bid is remembered
Dark areas in the retinal images, the dark areas includes angiosomes, hemorrhagic areas and dark noise region；
The angiosomes is determined from the enhancing image, and the region is removed from the dark space area image, is gone
Blood-vessel image；And
Determine the dark noise region from the enhancing image, and by the region from it is described remove blood-vessel image in remove, obtain
The hemorrhagic areas of the retinal images.
2. the method for claim 1, wherein the step of progress contrast enhancing to the retinal images is wrapped
Include：
The color value of the RGB triple channels of each pixel in the retinal images is normalized to the number between 0~1；
For each Color Channel in RGB, the color value of each pixel in enhancing image is determined according to below equation：
I1(x, y)=α I0(x,y)-β·I(x,y；δ)+γ
Wherein, I1(x, y) represents that the enhancing image coordinate is the color value of the pixel of (x, y), I0(x, y) represents to regard described
Coordinate is the color value of the pixel of (x, y), I (x, y in nethike embrane image；δ) represent that coordinate is (x, y) in the retinal images
Pixel local mean value, wherein, the local mean value is is drawn through the gaussian filtering that window size and variance are δ.
3. the method for claim 1, wherein described pair of enhancing image is filtered processing, to extract the view
The step of background image of film image, includes：
The multiple wave filters with different windows size of generation；
To tri- Color Channels of RGB of each pixel in the enhancing image, the multiple wave filter is respectively adopted each passage is entered
Row filtering process, obtains multiple filter results of each passage；And
Multiple filter results of each passage are taken into the average color value as the passage, so as to obtain the background image.
4. the method as described in claim 1 or 3, wherein, described to be filtered into Wiener filtering, its calculation formula is：
<mrow>
<mover>
<mi>F</mi>
<mo>^</mo>
</mover>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mo>&lsqb;</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mfrac>
<mrow>
<mo>|</mo>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
</mrow>
<mrow>
<mo>|</mo>
<mi>H</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
<mo>+</mo>
<mi>K</mi>
</mrow>
</mfrac>
<mo>&rsqb;</mo>
<mi>G</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
</mrow>
Wherein,The frequency-domain transform of the image extracted by Wiener filtering, G (u, v) is that Wiener filtering is presently in reason
The frequency-domain transform of image, H (u, v) is degenrate function, and K is fixed constant.
5. the method for claim 1, wherein the RGB color according to each pixel in the error image is worth to
The step of dark space area image, includes：
Obtain the color value of the RGB triple channels of each pixel in the error image, and according to each passage color value got come
Determine the color threshold of the passage；And
By contrasting the color threshold of each passage color value of each pixel and respective channel in the error image, by the pixel
Each passage color value mark be or mark be so that the error image is converted into dark space area image, the dark space
Area image is bianry image.
6. the method for claim 1, wherein described determine the angiosomes from the enhancing image, and should
The step of region is removed from the dark space area image includes：
The enhancing image is repeatedly filtered using multiple window sizes under different variances, respectively obtained under each variance
Multiple filter results, and the multiple filter result is averaged, obtain the filtering average under the variance；
Filtering average under each variance is merged, and row threshold division is entered to the image after merging, intermediate image is obtained, institute
Stating intermediate image includes pseudo- angiosomes and angiosomes, and it is bianry image；
The pseudo- angiosomes in the image is determined by carrying out connected domain analysis to the intermediate image；
The pseudo- angiosomes is removed from the intermediate image, the distribution map of the angiosomes is obtained, blood vessel point is designated as
Butut；And
By by the RGB color of respective pixel in the RGB color value of each pixel of dark space area image and the vascular distribution figure
Value takes difference, and the angiosomes is removed from the dark space area image.
7. method as claimed in claim 6, wherein, it is described to determine this by carrying out connected domain analysis to the intermediate image
The step of pseudo- angiosomes in image, includes：
Determine each connected domain in the intermediate image；
The property value of each connected domain is calculated, the property value includes the area, girth, the minimum comprising the connected domain of the connected domain
Rectangle frame, and have with region in oval eccentricity, long axis length and the minor axis length of identical standard second-order moment around mean
It is at least one；And
Judge whether meet the first predetermined condition between the property value of each connected domain, if so, being then pseudo- blood by the connected component labeling
Area under control domain.
8. a kind of retinal images hemorrhagic areas segmenting device, is performed in computing device, the device includes：
Image pre-processing unit, carries out contrast enhancing suitable for obtaining retinal images to be split, and to the image, obtains institute
The enhancing image of retinal images is stated, and processing is filtered to the enhancing image, to extract the retinal images
Background image；
Error image generation unit, suitable for by it is described enhancing image in each pixel RGB color value with it is right in the background image
Answer the RGB color value of pixel to take difference, obtain error image；
Dark areas determining unit, it is described suitable for being worth to dark space area image according to the RGB color of each pixel in the error image
The dark areas in the retinal images has been remembered in area image acceptance of the bid in dark space, the dark areas include angiosomes, hemorrhagic areas and
Dark noise region；
Blood vessel removal unit, suitable for determining the angiosomes from the enhancing image, and by the region from the dark areas
Removed in image, obtain blood-vessel image；And
Dark noise removal unit, suitable for determining the dark noise region from the enhancing image, and the region is gone from described
Removed in blood-vessel image, obtain the hemorrhagic areas of the retinal images.
9. a kind of computing device, including：
At least one processor；With
Have program stored therein the memory of instruction；
Wherein, the processor is configured as being performed as in claim 1-7 according to the programmed instruction stored in the memory
Method described in any one.
10. the program in a kind of computer-readable recording medium for the instruction that has program stored therein, the computer-readable recording medium
Instruction can be read by computing device so that method of the computing device as any one of claim 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710308401.7A CN107146231B (en) | 2017-05-04 | 2017-05-04 | Retinal image bleeding area segmentation method and device and computing equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710308401.7A CN107146231B (en) | 2017-05-04 | 2017-05-04 | Retinal image bleeding area segmentation method and device and computing equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107146231A true CN107146231A (en) | 2017-09-08 |
| CN107146231B CN107146231B (en) | 2020-08-07 |
Family
ID=59774023
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710308401.7A Active CN107146231B (en) | 2017-05-04 | 2017-05-04 | Retinal image bleeding area segmentation method and device and computing equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107146231B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108280816A (en) * | 2017-12-19 | 2018-07-13 | 维沃移动通信有限公司 | A kind of gaussian filtering method and mobile terminal |
| CN108596895A (en) * | 2018-04-26 | 2018-09-28 | 上海鹰瞳医疗科技有限公司 | Eye fundus image detection method based on machine learning, apparatus and system |
| CN108577803A (en) * | 2018-04-26 | 2018-09-28 | 上海鹰瞳医疗科技有限公司 | Eye fundus image detection method based on machine learning, apparatus and system |
| CN109993731A (en) * | 2019-03-22 | 2019-07-09 | 依未科技(北京)有限公司 | A kind of eyeground pathological changes analysis method and device |
| CN113822897A (en) * | 2021-11-22 | 2021-12-21 | 武汉楚精灵医疗科技有限公司 | Blood vessel segmentation method, terminal and computer-readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101520888B (en) * | 2008-02-27 | 2012-06-27 | 中国科学院自动化研究所 | Method for enhancing blood vessels in retinal images based on the directional field |
| CN104102899A (en) * | 2014-05-23 | 2014-10-15 | 首都医科大学附属北京同仁医院 | Retinal vessel recognition method and retinal vessel recognition device |
-
2017
- 2017-05-04 CN CN201710308401.7A patent/CN107146231B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101520888B (en) * | 2008-02-27 | 2012-06-27 | 中国科学院自动化研究所 | Method for enhancing blood vessels in retinal images based on the directional field |
| CN104102899A (en) * | 2014-05-23 | 2014-10-15 | 首都医科大学附属北京同仁医院 | Retinal vessel recognition method and retinal vessel recognition device |
Non-Patent Citations (2)
| Title |
|---|
| 吕卫 等: ""彩色眼底图像糖网渗出物的自动检测"", 《光电工程》 * |
| 吕菲，赵兴群: ""基于视频的舌下微循环血流灌注自动评价方法"", 《中国医疗器械杂志》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108280816A (en) * | 2017-12-19 | 2018-07-13 | 维沃移动通信有限公司 | A kind of gaussian filtering method and mobile terminal |
| CN108280816B (en) * | 2017-12-19 | 2020-09-18 | 维沃移动通信有限公司 | Gaussian filtering method and mobile terminal |
| CN108596895A (en) * | 2018-04-26 | 2018-09-28 | 上海鹰瞳医疗科技有限公司 | Eye fundus image detection method based on machine learning, apparatus and system |
| CN108577803A (en) * | 2018-04-26 | 2018-09-28 | 上海鹰瞳医疗科技有限公司 | Eye fundus image detection method based on machine learning, apparatus and system |
| CN108596895B (en) * | 2018-04-26 | 2020-07-28 | 上海鹰瞳医疗科技有限公司 | Fundus image detection method, device and system based on machine learning |
| CN109993731A (en) * | 2019-03-22 | 2019-07-09 | 依未科技(北京)有限公司 | A kind of eyeground pathological changes analysis method and device |
| CN113822897A (en) * | 2021-11-22 | 2021-12-21 | 武汉楚精灵医疗科技有限公司 | Blood vessel segmentation method, terminal and computer-readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107146231B (en) | 2020-08-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107146231A (en) | Retinal image bleeding area segmentation method and device and computing equipment | |
| CN107038704A (en) | Retina image exudation area segmentation method and device and computing equipment | |
| CN108198184B (en) | Method and system for vessel segmentation in contrast images | |
| CN107123124A (en) | Retina image analysis method and device and computing equipment | |
| CN114820654A (en) | Blood vessel segmentation method, blood vessel segmentation device, medical imaging equipment and storage medium | |
| CN109101994B (en) | Fundus image screening method and device, electronic equipment and storage medium | |
| CN108764342B (en) | Semantic segmentation method for optic discs and optic cups in fundus image | |
| CN110858399B (en) | Method and apparatus for providing post-examination images of a virtual tomographic stroke | |
| US11967181B2 (en) | Method and device for retinal image recognition, electronic equipment, and storage medium | |
| US11379989B2 (en) | Method and device of extracting label in medical image | |
| CN111242933B (en) | Retinal image artery and vein classification device, apparatus, and storage medium | |
| CN110443148A (en) | A kind of action identification method, system and storage medium | |
| CN109583364A (en) | Image-recognizing method and equipment | |
| CN111415304A (en) | Underwater vision enhancement method and device based on cascade deep network | |
| CN110473176B (en) | Image processing method and device, fundus image processing method and electronic equipment | |
| CN107133932A (en) | Retina image preprocessing method and device and computing equipment | |
| CN114399480A (en) | Method and device for detecting severity of vegetable leaf disease | |
| CN111881706A (en) | Living body detection, image classification and model training method, device, equipment and medium | |
| JP2018185265A (en) | Information processor, method for control, and program | |
| Vamsi et al. | Early Detection of Hemorrhagic Stroke Using a Lightweight Deep Learning Neural Network Model. | |
| CN116843647A (en) | Method and device for determining lung field area and evaluating lung development, electronic equipment and medium | |
| CN107038705A (en) | Retinal image bleeding area segmentation method and device and computing equipment | |
| CN112801238B (en) | Image classification method and device, electronic equipment and storage medium | |
| CN115731214A (en) | Medical image segmentation method and device based on artificial intelligence | |
| CN115131361A (en) | Training of target segmentation model, focus segmentation method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220511 Address after: 519000 unit Z, room 615, 6th floor, main building, No. 10, Keji 1st Road, Gangwan Avenue, Tangjiawan Town, Xiangzhou District, Zhuhai City, Guangdong Province (centralized office area) Patentee after: Zhuhai Quanyi Technology Co.,Ltd. Address before: 272500 No. 032, juntun Township commercial street, Wenshang County, Jining City, Shandong Province Patentee before: Ji Xin |