CN103246866A - Image processing method and device for correcting red eyes - Google Patents
Image processing method and device for correcting red eyes Download PDFInfo
- Publication number
- CN103246866A CN103246866A CN2012100262081A CN201210026208A CN103246866A CN 103246866 A CN103246866 A CN 103246866A CN 2012100262081 A CN2012100262081 A CN 2012100262081A CN 201210026208 A CN201210026208 A CN 201210026208A CN 103246866 A CN103246866 A CN 103246866A
- Authority
- CN
- China
- Prior art keywords
- region
- candidate region
- brightness
- interest
- color
- 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.)
- Pending
Links
Images
Abstract
The invention relates to an image processing method and device for correcting red eyes. The image processing method comprises the following steps: receiving and detecting a human face region in a to-be-processed image, setting at least one region of interest in the human face region, carrying out segmentation on the region of interest according to a color model to produce a plurality of candidate regions, screening the candidate regions respectively according to a candidate region filtering method, judging whether one color candidate region is produced after screening, if yes, calculating the brightness values of a plurality of pixels in the region of interest by a contrast mask, and carrying out segmentation on the region of interest according to the distribution of the brightness values calculated to produce the brightness candidate regions, taking the overlapped areas between the color candidate areas and the brightness candidate areas as red eye regions, and correcting the red eye areas to produce a corrected image.
Description
Technical field
The invention relates to a kind of image treatment method and device, and particularly relevant for a kind of image treatment method and device of revising blood-shot eye illness.
Background technology
The pupil that red eye phenomenon refers to eyes presents red phenomenon in photochrome.Main cause is because human eye pupil under the environment of darkness can amplify, and then increase light is injected amphiblestroid scope, so common people use flashlamp to take pictures, the high light of flashlamp will be penetrated blood capillary tissue arround the retina rear of eyes, and its red light that reflects just can cause the red eye phenomenon on the photochrome.
It is quite lofty and not attractive in appearance that red eye phenomenon seems on visual effect, not being that photographer is desired sees, therefore how to revise at the photo of red eye phenomenon, non-natural red eye phenomenon found out and it is modified to and approach the color of natural human eye originally, make it eliminate lofty sensation, real is an important topic in the image processing technique field.
Please refer to No. 7746385 patents of U.S. bulletin patent US, this patent disclosed a kind of will not use many captured previous images of flashlamp as many with reference to image, and with using flashlamp a captured present image is compared, revise the red eye phenomenon in the present image thus.Wherein, with many with reference to image with before at present image is compared, must earlier previous image be amplified and present image be dwindled, making previous image is consistent with the resolution of present image.In addition, because previous image is also inequality with the acquisition time of present image, cause image error for fear of situations such as hand shake or person taken picture move, also must carry out geometry correction (geometric alignment), last again at previous image and the difference place of image at present, determine the red eye phenomenon zone and revise.
Then, please refer to No. 7852377 patents of U.S. bulletin patent US, this patent has disclosed zone that a kind of utilization forms red pixel and has done how much and judge, mainly based on the circle differentiation, filter out the zone of border circular areas or sub-circular, and further above-mentioned zone is outwards expanded, simultaneously with reference to ambient conditions, pick out the red eye region of desire correction according to this.
Yet No. 7746385 patents of U.S. bulletin patent US are owing to need earlier through adjusting flow processs such as resolution, geometry correction, and the high and required resource of computational complexity is higher, for example because of many of needs accesses with reference to image, so the quantity of impact damper and space are all bigger.And its range of application is more restricted, if photographer directly utilize flashlamp to take pictures to there is no can reference image not for a flash then can't use the method.In addition, since No. 7852377 patents of U.S. bulletin patent US mainly with border circular areas as judgement, yet, in the photo of actual generation red eye phenomenon, eye shape differs, and to establish a capital be the circle of standard, for instance, and the eyes of semi-closure, the eyes of stravismus etc., face the person taken picture is not under the situation of phtographic lens, may produce the situation of erroneous judgement (false positive).
Summary of the invention
In view of this, the invention provides a kind of image treatment method of revising blood-shot eye illness, can effectively reduce the erroneous judgement probability, and mark the red eye region of desire correction more accurately, and then revise automatically.
The invention provides a kind of image processor of revising blood-shot eye illness, can be directly to one pick-up image detect, mark the red eye region of desire correction fast and revise automatically, and then export image after the correction.
The present invention proposes a kind of image treatment method of revising blood-shot eye illness, and it comprises the following steps.Elder generation receives a pending image, and detects the human face region in this pending image, and in this person's face zone, set one or more region-of-interest (Region of Interest, ROI).Then, according to colour model above-mentioned region-of-interest is carried out mark (segmentation) and produce a plurality of candidate regions thus.And respectively each candidate region is screened according to a candidate region filtration method, whether produce a color candidate region after the judgement screening according to this.If, then utilize a contrast shade (contrast mask) that the brightness value of a plurality of pixels in the above-mentioned region-of-interest is carried out computing, and utilize the brightness value distribution after the computing to carry out mark, produce the brightness candidate region according to this.At last,, and red eye region revised to produce thus revise the back image as red eye region according to color candidate region and overlapping part.
In one embodiment of this invention, wherein if do not produce the color candidate region after judging screening, above-mentioned image treatment method also comprises according to the candidate region filtration method brightness candidate region is screened, and the brightness candidate region after directly will screening is revised red eye region thus as red eye region.
In one embodiment of this invention, above-mentioned utilization contrast shade carries out mark to one or more region-of-interest and also comprises the following steps.Utilize earlier the contrast shade that the brightness value of a plurality of pixels in the above-mentioned region-of-interest is carried out computing and produce a plurality of responses according to this.Orient reference center's point according to these responses again.And the brightness value according to a plurality of pixels of the adjacent domain of this reference center's point carries out computing, produces meta brightness value and standard deviation thus.Afterwards, according to meta brightness value and the formed brightness value distributed area of standard deviation above-mentioned region-of-interest is carried out mark, produce the brightness candidate region according to this.
In one embodiment of this invention, when the above-mentioned step of orienting reference center's point thus according to these responses comprises peak response value in producing these responses, select the central point of contrast shade to correspond to position in the above-mentioned region-of-interest as reference center's point.
In one embodiment of this invention, above-mentioned each candidate region the screening according to the candidate region filtration method comprises the following steps.Center with each candidate region is the center of circle earlier, and forms first border circular areas with first predeterminable range thus as radius, utilizes first border circular areas to carry out computing with interior chroma value with first border circular areas a plurality of pixels in addition, produces first eigenwert thus.Judge that then whether first eigenwert is greater than threshold value.If first eigenwert is greater than threshold value, the candidate region is judged to be the color candidate region.
In one embodiment of this invention, above-mentioned image treatment method comprises that also elder generation is the center of circle with the center of each candidate region, and with second predeterminable range as radius to form second border circular areas, utilize second border circular areas to carry out computing with interior chroma value with second border circular areas a plurality of pixels in addition, and then produce second eigenwert.And then judge that whether first eigenwert or second eigenwert are greater than threshold value.If one of them is greater than threshold value at least for first eigenwert or second eigenwert, then the candidate region is judged to be the color candidate region.
In one embodiment of this invention, above-mentioned image treatment method also is included in the human face region, analyzes the relative position relation between each candidate region, filters out the color candidate region thus.
The present invention proposes a kind of image processor of revising blood-shot eye illness in addition, and it comprises human face detection module, color indicia module, filtering module, contrast shade module and red-eye correction module.Wherein, the human face detection module is in order to receiving a pending image, and detects the human face region in the pending image thus, and sets one or more region-of-interest in human face region.The color indicia module is coupled to the human face detection module, utilizes colour model that above-mentioned region-of-interest is carried out mark and produces a plurality of candidate regions thus.Filtering module is coupled to the color indicia module, respectively each candidate region is screened, and whether produces the color candidate region after judging screening.Contrast shade module is coupled to filtering module, if filtering module produces the color candidate region really after screening, then contrasting the shade module utilizes brightness value after a contrast shade carries out computing and utilizes computing the brightness value of a plurality of pixels in the above-mentioned region-of-interest to distribute to carry out mark, produce the brightness candidate region according to this.The red-eye correction module is coupled to filtering module and contrast shade module, according to the overlapping part in color candidate region and brightness candidate region as red eye region, and to this red eye region revise produce thus revise after image.
In one embodiment of this invention, above-mentioned filtering module does not produce the color candidate region after each candidate region is screened, and then filtering module further screens the brightness candidate region, to produce the brightness candidate region after screening.Brightness candidate region after the red-eye correction module directly will be screened is as red eye region, and this red eye region is revised.
In one embodiment of this invention, above-mentioned contrast shade module comprises positioning unit, arithmetic element and brightness indexing unit.Wherein, positioning unit utilization contrast shade carries out computing to the brightness value of a plurality of pixels in the above-mentioned region-of-interest and produces a plurality of responses according to this, and orients reference center's point according to these responses.Arithmetic element is coupled to positioning unit, and the brightness value of a plurality of pixels of the adjacent domain of benchmark central point carries out computing, produces meta brightness value and standard deviation thus.The brightness indexing unit is coupled to arithmetic element, distributes according to meta brightness value and the formed brightness value of standard deviation above-mentioned region-of-interest is carried out mark, produces the brightness candidate region according to this.
In one embodiment of this invention, above-mentioned positioning unit selects the central point of contrast shade to correspond to position at least one region-of-interest as reference center's point when producing the peak response value.
In one embodiment of this invention, above-mentioned filtering module is the center of circle with the center of each candidate region that the color indicia module is produced, and utilizes first predeterminable range to form first border circular areas as radius.In addition, filtering module and utilize first border circular areas to carry out computing to produce first eigenwert with the chroma values of a plurality of pixels beyond interior and first border circular areas.Filtering module judges that first eigenwert is the color candidate region greater than the candidate region of a threshold value.
In one embodiment of this invention, above-mentioned filtering module forms second border circular areas with second predeterminable range as radius, and filtering module also utilizes second border circular areas to carry out computing to produce second eigenwert with interior chroma value with second border circular areas a plurality of pixels in addition.Filtering module judge first eigenwert or second eigenwert at least one of them candidate region greater than threshold value be the color candidate region.
In one embodiment of this invention, the relative position relation between the candidate region that above-mentioned filtering module analysis color indicia module produces, filtering module filters out the color candidate region according to analysis result.
Based on above-mentioned, image treatment method and the device of correction blood-shot eye illness provided by the present invention, adopt a contrast shade to carry out the resulting brightness value distributed area of computing pending image is carried out mark, and then generation brightness candidate region, be aided with the red eye region that the color candidate region determines the desire compensation jointly, so can increase accuracy and tolerance that red eye region is differentiated.
For above-mentioned feature and advantage of the present invention can be become apparent, embodiment cited below particularly, and cooperate appended accompanying drawing to be described in detail below.
Description of drawings
Fig. 1 is the calcspar of the image processor of the correction blood-shot eye illness that illustrates according to one embodiment of the invention.
Fig. 2 is the process flow diagram of the image treatment method of the correction blood-shot eye illness that illustrates according to one embodiment of the invention.
Fig. 3 is the calcspar of the image processor that illustrates according to another embodiment of the present invention.
Fig. 4 is the process flow diagram of the image treatment method of the correction blood-shot eye illness that illustrates according to another embodiment of the present invention.
Fig. 5 A is the contrast shade that illustrates according to another embodiment of the present invention and the synoptic diagram of region-of-interest.
The synoptic diagram of contrast shade position when Fig. 5 B is the generation peak response value that illustrates according to another embodiment of the present invention.
Fig. 5 C is the enlarged diagram of a plurality of pixels of reference center's point C among Fig. 5 B and adjacent domain.
Fig. 6 A and Fig. 6 B are the center of the candidate region that illustrates according to yet another embodiment of the invention and the synoptic diagram of radius.
[main element symbol description]
100,300: image processor
110: the human face detection module
120: the color indicia module
130: filtering module
140: contrast shade module
142: positioning unit
144: arithmetic element
146: the brightness indexing unit
150: the red-eye correction module
500: pending image
501,503: region-of-interest
505: the contrast shade
610,620: circular candidate region
612,622: border circular areas
C: reference center's point
D1, d2, dn: the direction of arrow
P1～P9: pixel
L1～L4: anchor point
R, ra, rb: radius
S210～S260: each step of revising the image treatment method of blood-shot eye illness
S410～S460: each step of revising the image treatment method of blood-shot eye illness.
Embodiment
The present invention proposes a kind of image treatment method and device of revising blood-shot eye illness, can directly handle a pending image, and do not need to utilize many to compare with reference to image.In addition, the present invention utilizes the technology of contrast shade (contrast mask) to be aided with and filters out the candidate region that produces red eye phenomenon more accurately.No matter and whether the zone that produces red eye phenomenon be the circle of standard, the present invention all can detect and then reduce erroneous judgement (false positive) probability effectively.In order to make content of the present invention more clear, below enumerate the example that embodiment can implement really according to this as the present invention.The embodiment that proposes as the usefulness of explanation, is not to limit interest field of the present invention only.
Fig. 1 is the calcspar of the image processor of the correction blood-shot eye illness that illustrates according to one embodiment of the invention.Please refer to Fig. 1, the image processor 100 of present embodiment for example is intelligent mobile phone with IPF of digital camera, slr camera, digital camera or other, panel computer etc., is not limited to above-mentioned.Image processor 100 comprises human face detection module 110, color indicia module 120, filtering module 130, contrast shade module 140 and red-eye correction module 150.Each above-mentioned module can be the functional module that hardware and/or software are realized.Wherein hardware can comprise that central processing unit, chipset, microprocessor etc. have the combination of hardware device or the above-mentioned hardware device of calculation function, and software then can be driver, application program, operating system etc.
Fig. 2 is the process flow diagram of the image treatment method of the correction blood-shot eye illness that illustrates according to one embodiment of the invention.The method of present embodiment is applicable to the image processor 100 of Fig. 1, below namely the arrange in pairs or groups detailed step of each the module declaration present embodiment image treatment method in the image processor 100:
At first, in step S210, human face detection module 110 elder generations receive a pending image, and detect the human face region in this pending image, and in this person's face zone, set one or more region-of-interest (Region of Interest, ROI).This step focuses on search area is dwindled, and reduces the time of the required cost of image processing thus.Because the zone that red eye phenomenon produces system betides the eyes position, that is to say, region-of-interest must comprise eyes position and adjacent domain on every side thereof, therefore human face detection module 110 is after detecting human face region, can utilize fast searching (quick search) method at setting one or more region-of-interest in the human face region, wherein the area of region-of-interest and number then can be done design according to actual presentation content, do not limit at this.
Then, in step S220, color indicia module 120 is carried out mark (segmentation) according to colour model (color model) to each region-of-interest and is produced a plurality of candidate regions thus.Colour model for example is primaries (RGB) model.Because the zone that red eye phenomenon produces is generally the zone of red pixel clustering, red pixel refers to be in the great majority with red composition in the color composition of pixel.Therefore, color indicia module 120 for example definable goes out a red interval range in the RGB model, and each pixel in each region-of-interest is carried out mark.For instance, color indicia module 120 can all be labeled as 1 with the pixel that belongs to red interval range in the region-of-interest, and remaining pixel is labeled as 0., just the pixel circle that belongs to red interval range can be elected after all mark is finished in each pixel, so just can produce a plurality of candidate regions.
After producing a plurality of candidate regions, but just subsequent steps S230, and filtering module 130 screens each candidate region respectively according to a candidate region filtration method, whether produces a color candidate region after judging screening according to this.Because the candidate region that produces in step S220 is except the pupil part that produces red eye phenomenon, also might be to be positioned at the unique point that other red pixels such as canthus, the corners of the mouth are assembled, this kind partly also is made as the situation that the candidate region is erroneous judgement (false positive) with non-pupil, will cause correcting mistakes, extent of amendment is too much or very few.So, in step S230, will do further a plurality of candidate regions according to a candidate region filtration method and differentiate and screening, the candidate region that stays by screening conditions is referred to herein as the color candidate region.The content of detailed candidate region filtration method will in after embodiment be elaborated.Wherein, if judge the existence of color candidate region, then subsequent steps S240 are arranged really; There are not color candidate region, then subsequent steps S260 if judge.
In step S240, contrast shade module 140 utilizes contrast shade (contrast mask) that the brightness value of a plurality of pixels in each region-of-interest is carried out computing, and utilize the brightness value distributed area after the computing to carry out mark, produce the brightness candidate region according to this.In the present embodiment, brightness value refers to pixel when encoding with the yuv format of color space, wherein the pixel value of Y passage (Y channel).Therefore, region-of-interest a plurality of pixels are wherein carried out after the computing, can obtain a brightness value distributed area, then, facility is come region-of-interest is detected with this brightness value distributed area, all pixels that the brightness value of pixel belongs in this Luminance Distribution interval are carried out mark, and the pixel circle that all are labeled elected to obtain the brightness candidate region.
Then, in step S250, then red-eye correction module 150 compares the previous color candidate region that filters out and brightness candidate region, and the part that both are overlapping is as red eye region, at last this red eye region is revised, just can be produced one and revise the back image.
Yet, if before do not filtered out the color candidate region, the brightness candidate region that then directly utilizes step S240 to produce adopts back favored area filtration method to do screening further, and the brightness candidate region that directly will meet screening conditions is as red eye region, equally this candidate region is revised, and can be obtained revising the back image.
The present invention also filters the color candidate region that produces except utilizing colour model (the red composition of pixel) to carry out mark, also adopt a contrast shade to carry out mark, and then produce the brightness candidate region, determine best red eye region jointly with color candidate region and brightness candidate region.If only carry out mark with colour model, for the red eye phenomenon of low composition, for example rufous etc. can't screen, yet, do by the contrast shade and to differentiate the red eye phenomenon that can detect low red composition further.Accordingly, the present invention utilizes the contrast shade can assist a ruler in governing a country the differentiation result of colour model, has increased accuracy and tolerance that red eye region is differentiated.
The following example that can implement according to this really as the present invention for an embodiment in addition.Fig. 3 is the calcspar of the image processor that illustrates according to another embodiment of the present invention.What must illustrate is that Fig. 3 is wherein a kind of embodiment of the image processor 100 of Fig. 1.Please refer to Fig. 3, the brightness indexing unit 146 that contrast shade module 140 comprises positioning unit 142, is coupled to the arithmetic element 144 of positioning unit and is coupled to arithmetic element.
Fig. 4 is the process flow diagram of the image treatment method of the correction blood-shot eye illness that illustrates according to another embodiment of the present invention, and wherein, Fig. 4 is a kind of process flow diagram of detailed embodiment of image treatment method of the correction blood-shot eye illness of Fig. 2.Below the method for operation of image processor 300 will be described with Fig. 4.Please be simultaneously with reference to Fig. 3 and Fig. 4.
At first, human face detection module 110 receives a pending image earlier, and detects the human face region in this pending image, and sets one or more region-of-interest (step S410) in this person's face zone.Then, color indicia module 120 is carried out mark according to colour model to each region-of-interest and is produced a plurality of candidate regions (step S420) thus.Afterwards, filtering module 130 screens each candidate region respectively according to a candidate region filtration method, whether produces a color candidate region (step S430) after the judgement screening according to this.Above-mentioned step S410～S430, does not give unnecessary details at this so its implementation detail is described in last embodiment the same as or similar to step S210～S230.
Next, the utilization of the step S440 contrast shade step of each region-of-interest being carried out mark and producing the brightness candidate region is according to this then finished by step S442～S448.
In step S442, positioning unit 142 utilizes the contrast shade respectively the brightness value of a plurality of pixels in each region-of-interest to be carried out computing earlier, produces a plurality of responses (response) value according to this.Fig. 5 A is the contrast shade that illustrates according to another embodiment of the present invention and the synoptic diagram of region-of-interest.Please refer to Fig. 5 A, in pending image 500, for example comprise two region-of- interests 501 and 503, positioning unit 142 utilizes contrast shade 505 to scan in region-of-interest 501, meaning namely in regular turn by direction of arrow d1, d2 ..., dn carries out computing from upper left to the bottom right, wherein, the scope that positioning unit 142 utilizes contrast shade 505 to cover in region-of-interest 501 is chosen partly, and the brightness value of pixel carries out computing and then produces a response, wherein, more more big near the resulting response in eyes position.
Then in step S444, positioning unit 142 is just oriented reference center's point according to a plurality of responses that produce in region-of-interest 501.In detail, positioning unit 142 is chosen the peak response value in a plurality of responses earlier, and finds out the position of contrast shade when producing this peak response value.The synoptic diagram of contrast shade position when for instance, Fig. 5 B is the generation peak response value that illustrates according to another embodiment of the present invention.Please refer to Fig. 5 B, C is put as reference center in the position that the central point of selection contrast shade 505 corresponds in the region-of-interest 501.
Next in step S446, arithmetic element 144 is carried out computing according to the brightness value of a plurality of pixels of the adjacent domain of this reference center's point C, produces meta brightness value M and standard deviation S thus.For instance, Fig. 5 C is the enlarged diagram of a plurality of pixels of reference center's point C among Fig. 5 B and adjacent domain.Please refer to Fig. 5 C, utilize the brightness value of pixel P1～P9 (wherein pixel P5 is reference center's point C), just can calculate a meta brightness value M.Wherein, the choosing of the neighborhood pixels shown in Fig. 5 C (that is, pixel P1～P9) only is an example embodiment, and the present invention is not limited to this.Then, the neighborhood pixels according to meta brightness value M and reference center's point C periphery compares and then produces standard deviation S again, and wherein, the neighborhood pixels of periphery for example is centered by reference center's point C, a plurality of pixels on the transverse axis X.
In step S448,146 of brightness indexing units carry out mark according to meta brightness value M and the formed brightness value distributed area of standard deviation S to region-of-interest 501, produce the brightness candidate region according to this.
At last, red-eye correction module 150 compares the previous color candidate region that filters out and brightness candidate region, and the part that both are overlapping is revised this red eye region at last as red eye region, just can produce one and revise back image (step S450).Yet, if before do not filtered out the color candidate region, the brightness candidate region that then directly utilizes step S440 to produce adopts back favored area filtration method to do screening further, and the brightness candidate region that directly will meet screening conditions is as red eye region, red-eye correction module 150 is revised this candidate region equally, and can obtain revising back image (step S460).
Describe the step that the candidate region filtration method screens each candidate region in detail for an embodiment again with next.What must illustrate is, if candidate region itself is the candidate region that marks via colour model, then by can being judged to be the color candidate region after the screening of candidate region filtration method, and can be further with the brightness candidate region carry out overlap ratio to after determine the red eye region of desire compensation again.If candidate region itself is the candidate region that marks via the contrast shade, then by can directly being judged to be red eye region after the screening of candidate region filtration method.
Fig. 6 A and Fig. 6 B are the center of the candidate region that illustrates according to yet another embodiment of the invention and the synoptic diagram of radius.Wherein, each step of present embodiment can be finished by the filtering module 130 of Fig. 1 or Fig. 3.
Please refer to Fig. 6 A and Fig. 6 B, what must illustrate is, the candidate region must be earlier handled (morphotogical process) and then is the center of circle with the anchor point L1 of candidate region via the simple geometric form, and being that radius forms a circular candidate region 610 and 620 apart from r between anchor point L1 and the anchor point L2.
As shown in Figure 6A, again with the first predeterminable range ra (namely, distance between anchor point L1 and the anchor point L3) forms first border circular areas 612 thus as radius, utilize first border circular areas 612 to carry out computing with interior chroma value with first border circular areas 612 a plurality of pixels in addition, produce first eigenwert thus.In detail, in the present embodiment, the chroma value for example is pixel when encoding with the yuv format of color space, wherein the pixel value of V passage (V channel).Utilize first border circular areas 612 with the difference of the chroma value of a plurality of pixels beyond the chroma value of interior a plurality of pixels and first border circular areas 612 as first eigenwert.Then and judge that whether first eigenwert is greater than threshold value.If first eigenwert greater than threshold value, then represents the candidate region and passes through screening conditions.
Shown in Fig. 6 B, again with the second predeterminable range rb (namely, distance between anchor point L1 and the anchor point L4) forms second border circular areas 622 thus as radius, utilize second border circular areas 622 to carry out computing with interior chroma value with second border circular areas 622 a plurality of pixels in addition, produce second eigenwert thus.Then and judge that whether second eigenwert is greater than threshold value.If second eigenwert greater than threshold value, then represents the candidate region and passes through screening conditions.
What must illustrate is that the candidate region filtration method of present embodiment can judge in regular turn that whether first eigenwert or second eigenwert are greater than threshold value.If one of them then represents the candidate region and passes through screening conditions greater than threshold value at least for first eigenwert or second eigenwert.In addition, in order to increase the accuracy of detection, can repeatedly adjust the distributing position of pixel sampling spot, perhaps confirm repeatedly as radius according to different predeterminable ranges.Accordingly, even the ratio difference that the zone that red eye phenomenon produces accounts in the pupil can distinguish still whether it is red eye region, for example blood-shot eye illness may take whole pupil region, perhaps sees red because of the eyes semi-closure and only accounts for about half pupil region etc.At last, if the candidate region is confirmed all can't to represent this candidate region and should be classified as erroneous judgement by the test of candidate region filtration method according to different screening conditions, meaning is that this candidate region is not to be positioned at red eye region.
In sum, the present invention can directly handle a pending image, and does not need to utilize many to compare with reference to image as prior art, can save system resource and computational complexity thus.In addition, the present invention also adopts a contrast shade to carry out the resulting brightness value distributed area of computing pending image is carried out mark, and then generation brightness candidate region, be aided with the red eye region that the color candidate region determines the desire compensation jointly, red eye region for low red composition still can detect by the present invention, therefore, the present invention can increase accuracy and the tolerance that red eye region is differentiated.Moreover the present invention has also proposed detection method more accurately for the detection of candidate region, even the ratio difference that the zone that red eye phenomenon produces accounts in the pupil can distinguish still whether it is red eye region.
Though the present invention discloses as above with embodiment; right its is not in order to limiting the present invention, any those skilled in the art, without departing from the spirit and scope of the present invention; when doing a little change and retouching, so protection scope of the present invention is as the criterion when looking the claim person of defining.
Claims (14)
1. image treatment method of revising blood-shot eye illness comprises:
Receive a pending image, detect the human face region in this pending image, and in this human face region, set at least one region-of-interest;
According to a color model this at least one region-of-interest is carried out mark and produce a plurality of candidate regions thus;
Respectively this candidate region is respectively screened according to a candidate region filtration method, whether produce a color candidate region after judging screening according to this;
If, utilize a brightness value after a contrast shade carries out computing and utilizes computing the brightness value of a plurality of pixels in this at least one region-of-interest to distribute this at least one region-of-interest is carried out mark, produce a brightness candidate region according to this; And
, and this red eye region revised produce one thus and revise the back image as a red eye region according to the overlapping part in this color candidate region and this brightness candidate region.
2. the image treatment method of correction blood-shot eye illness according to claim 1 does not wherein produce this color candidate region after judging screening, and this image treatment method also comprises:
According to this candidate region filtration method this brightness candidate region is screened, and this brightness candidate region after directly will screening is revised this red eye region thus as this red eye region.
3. the image treatment method of correction according to claim 1 blood-shot eye illness, wherein utilize this contrast shade that this at least one region-of-interest is carried out mark and comprise with the step that produces this brightness candidate region:
Utilize this contrast shade that the brightness value of a plurality of pixels in this at least one region-of-interest is carried out computing to produce a plurality of responses;
The described response of foundation is to orient reference center point;
Brightness value to a plurality of pixels of the adjacent domain of this reference center's point carries out computing, to produce a meta brightness value and a standard deviation; And
According to this meta brightness value and formed this brightness value distribution of this standard deviation this at least one region-of-interest is carried out mark, produce this brightness candidate region according to this.
4. the image treatment method of correction according to claim 3 blood-shot eye illness wherein comprises with the step of orienting this reference center's point according to described response:
During a peak response value in producing described response, select the central point of this contrast shade to correspond to position in this at least one region-of-interest as this reference center's point.
5. the image treatment method of correction according to claim 1 blood-shot eye illness, wherein the step of each described candidate region being screened according to this candidate region filtration method comprises:
Center with this candidate region is the center of circle, and with one first predeterminable range as radius to form one first border circular areas, utilize this first border circular areas to carry out computing to produce one first eigenwert with the chroma values of a plurality of pixels beyond interior and this first border circular areas;
Judge that whether this first eigenwert is greater than a threshold value; And
If this first eigenwert is greater than this threshold value, this candidate region is this color candidate region.
6. the image treatment method of correction according to claim 5 blood-shot eye illness also comprises:
Center with this candidate region is the center of circle, and with one second predeterminable range as radius to form one second border circular areas, utilize this second border circular areas to carry out computing to produce one second eigenwert with the chroma values of a plurality of pixels beyond interior and this second border circular areas;
Judge that whether this first eigenwert or this second eigenwert are greater than this threshold value; And
If one of them is greater than this threshold value at least for this first eigenwert or this second eigenwert, this candidate region is this color candidate region.
7. the image treatment method of correction according to claim 1 blood-shot eye illness also comprises:
In this human face region, analyze the relative position relation between the described candidate region, filter out this color candidate region thus.
8. image processor of revising blood-shot eye illness comprises:
One human face detection module receives a pending image, detecting the human face region in this pending image, and sets at least one region-of-interest in this human face region;
One color mark module couples this human face detection module, utilizes a color model that this at least one region-of-interest is carried out mark to produce a plurality of candidate regions;
One filtering module couples this color indicia module, respectively this candidate region is respectively screened, and whether produces a color candidate region after judging screening;
One contrast shade module, couple this filtering module, if this filtering module produces this color candidate region, this contrast shade module utilizes a contrast shade that the brightness value of a plurality of pixels in this at least one region-of-interest is carried out computing, and utilize the brightness value distribution after the computing that this at least one region-of-interest is carried out mark, to produce a brightness candidate region; And
One red-eye correction module couples this filtering module and this contrast shade module,, and this red eye region correction is produced one revises the back image as a red eye region according to the overlapping part in this color candidate region and this brightness candidate region.
9. the image processor of correction according to claim 8 blood-shot eye illness, wherein:
This filtering module does not produce this color candidate region to after respectively screen this candidate region, this filtering module screens and produces this brightness candidate region after the screening to this brightness candidate region, and this brightness candidate region after this red-eye correction module directly will be screened is as this red eye region, and this red eye region is revised.
10. the image processor of correction according to claim 8 blood-shot eye illness, wherein this contrast shade module comprises:
Locating unit utilizes this contrast shade that the brightness value of a plurality of pixels in this at least one region-of-interest is carried out computing producing a plurality of responses, and according to described response to orient reference center point;
One arithmetic element couples this positioning unit, the brightness value of a plurality of pixels of the adjacent domain of this reference center's point is carried out computing, to produce a meta brightness value and a standard deviation; And
One brightness indexing unit couples this arithmetic element, distributes according to this meta brightness value and formed this brightness value of this standard deviation this at least one region-of-interest is carried out mark, to produce this brightness candidate region.
11. the image processor of correction blood-shot eye illness according to claim 10, wherein:
During the peak response value of this positioning unit in producing described response, select the central point of this contrast shade to correspond to position in this at least one region-of-interest as this reference center's point.
12. the image processor of correction blood-shot eye illness according to claim 8, wherein:
This filtering module is the center of circle with the center of respectively this candidate region that this color indicia module is produced, and utilize one first predeterminable range to form one first border circular areas as radius, this filtering module also utilizes this first border circular areas to carry out computing to produce one first eigenwert with interior chroma value with this first border circular areas a plurality of pixels in addition, and this filtering module judges that this first eigenwert is this color candidate region greater than this candidate region of a threshold value.
13. the image processor of correction blood-shot eye illness according to claim 12, wherein:
This filtering module forms one second border circular areas with one second predeterminable range as radius, this filtering module also utilizes this second border circular areas to carry out computing producing one second eigenwert with the chroma values of a plurality of pixels beyond interior and this second border circular areas, this filtering module judge this first eigenwert or this second eigenwert at least one of them this candidate region greater than this threshold value be this color candidate region.
14. the image processor of correction blood-shot eye illness according to claim 8, wherein:
This filtering module is analyzed the relative position relation between the described candidate region that this color indicia module produces, and this filtering module filters out this color candidate region according to analysis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012100262081A CN103246866A (en) | 2012-02-07 | 2012-02-07 | Image processing method and device for correcting red eyes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012100262081A CN103246866A (en) | 2012-02-07 | 2012-02-07 | Image processing method and device for correcting red eyes |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103246866A true CN103246866A (en) | 2013-08-14 |
Family
ID=48926379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012100262081A Pending CN103246866A (en) | 2012-02-07 | 2012-02-07 | Image processing method and device for correcting red eyes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103246866A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050232481A1 (en) * | 2004-04-16 | 2005-10-20 | Donghui Wu | Automatic red eye removal |
| US20060017825A1 (en) * | 2004-06-30 | 2006-01-26 | Khageshwar Thakur | Method and apparatus for effecting automatic red eye reduction |
| US20060056687A1 (en) * | 2004-09-15 | 2006-03-16 | Jonathan Brandt | Hierarchically locating a feature in a digital image |
| CN101489051A (en) * | 2008-01-17 | 2009-07-22 | 佳能株式会社 | Image processing apparatus and image processing method and image capturing apparatus |
| CN101620679A (en) * | 2009-07-22 | 2010-01-06 | 凌阳电通科技股份有限公司 | Method for eliminating red eye in image |
-
2012
- 2012-02-07 CN CN2012100262081A patent/CN103246866A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050232481A1 (en) * | 2004-04-16 | 2005-10-20 | Donghui Wu | Automatic red eye removal |
| US20060017825A1 (en) * | 2004-06-30 | 2006-01-26 | Khageshwar Thakur | Method and apparatus for effecting automatic red eye reduction |
| US20060056687A1 (en) * | 2004-09-15 | 2006-03-16 | Jonathan Brandt | Hierarchically locating a feature in a digital image |
| CN101489051A (en) * | 2008-01-17 | 2009-07-22 | 佳能株式会社 | Image processing apparatus and image processing method and image capturing apparatus |
| CN101620679A (en) * | 2009-07-22 | 2010-01-06 | 凌阳电通科技股份有限公司 | Method for eliminating red eye in image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP3298072B2 (en) | Video camera system | |
| US7444017B2 (en) | Detecting irises and pupils in images of humans | |
| US8055016B2 (en) | Apparatus and method for normalizing face image used for detecting drowsy driving | |
| US8773464B2 (en) | Methods and systems for collaborative-writing-surface image formation | |
| CN101443791B (en) | Improved foreground/background separation in digitl images | |
| CN104853668A (en) | Tiled image based scanning for head position for eye and gaze tracking | |
| CN101599175B (en) | Detection method for determining alteration of shooting background and image processing device | |
| CN106604005B (en) | A kind of projection TV Atomatic focusing method and system | |
| CN103997598A (en) | Method of tracking object using camera and camera system for object tracking | |
| JPH0944685A (en) | Face image processor | |
| CN103399695B (en) | Quadrangle frame identification method and device for intelligent wireless communication terminal | |
| KR101885728B1 (en) | Image stitching system, method and computer readable recording medium | |
| JP2009268085A (en) | Image trimming device and program | |
| CN100492088C (en) | Automatic focusing method | |
| JP5640622B2 (en) | Method for classifying red-eye object candidates, computer-readable medium, and image processing apparatus | |
| CN110290349A (en) | The method of the sitting posture state of lamps and lanterns and detecting user | |
| CN108572734A (en) | A kind of gestural control system based on infrared laser associated image | |
| CN109035331A (en) | A kind of aligning method and apparatus of signal lamp group | |
| CN111669492A (en) | Method for processing shot digital image by terminal and terminal | |
| CN1567377B (en) | Method for treating red-eye of digital image | |
| CN109003246A (en) | Eye repairs graph parameter detection method | |
| US10609275B2 (en) | Image processing device, image processing method, and recording medium | |
| CN103246866A (en) | Image processing method and device for correcting red eyes | |
| TW201328312A (en) | Image processing method and device for redeye correction | |
| CN115619636A (en) | Image stitching method, electronic device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20130814 |