CN101359364A - Image recognition method and image recognition device - Google Patents
Image recognition method and image recognition device Download PDFInfo
- Publication number
- CN101359364A CN101359364A CNA2007101382836A CN200710138283A CN101359364A CN 101359364 A CN101359364 A CN 101359364A CN A2007101382836 A CNA2007101382836 A CN A2007101382836A CN 200710138283 A CN200710138283 A CN 200710138283A CN 101359364 A CN101359364 A CN 101359364A
- Authority
- CN
- China
- Prior art keywords
- group
- image
- discrete cosine
- texture
- cosine transform
- 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
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
Abstract
The invention provides an image identification method. Firstly, the method judges if a group of discrete cosine conversion coefficients corresponding to an image, and/or a group of texture parameters corresponding to the group of discrete cosine conversion coefficients exist(s). If the judgment result is false, the method selectively processes a discrete cosine conversion or a reverse discrete cosine conversion towards the image according to a format of the image to generate the group of discrete cosine conversion coefficients, and generates the group of texture parameters according to the group of discrete cosine conversion coefficients. Then, the method compares the group of texture parameters with the target texture parameters to generate an identification result.
Description
Technical field
The present invention relates to image processing techniques, relate to a kind of method and device of image recognition in particular.
Background technology
In recent years, along with image science and technology is constantly progressive, the technology that digital picture is relevant also develops quite rapidly.Nowadays most photograph has all adopted digitized mode to take (as digital camera) and has stored (as storage card, CD or hard disk) with digitized form.Therefore the recognition technology of digital picture has the potentiality of its development owing to being widely used in monitoring management system and saving from damage in the burglary-resisting system.The employed digital picture recognition technology of tradition is at carrying out calculation process between image and the image mostly, utilizes pixel that the comparison of pixel is come the discriminating digit image.Yet the recognition speed of this kind digital picture recognition technology is quite slow and efficient is very low, supposes to have the huge digital picture of quantity to need identification, more can expend the work that a large amount of time and system resource are discerned.
Therefore, main category of the present invention is to provide a kind of image-recognizing method and device, to address the above problem.
Summary of the invention
A specific embodiment according to the present invention is a kind of image-recognizing method.In this specific embodiment, at first, this method judges whether to exist one group of discrete cosine transform corresponding to an image (DiscreteCosine Transform, DCT) coefficient and/or corresponding to one group of parametric texture of this group discrete cosine transform coefficient.If judged result is for denying, this method is according to a form of this image, optionally carry out a discrete cosine transform or an inverse discrete cosine transform (Inverse Discrete CosineTransform at this image, IDCT), producing this group discrete cosine transform coefficient, and produce this group parametric texture according to this group discrete cosine transform coefficient.Then, this method relatively should be organized parametric texture and one group of target texture parameter, to produce a recognition result.
Because the discrete cosine transform coefficient of this image is corresponding to its textural characteristics, therefore above-mentioned image-recognizing method can come these image texture features of comparison by discrete cosine transform coefficient that relatively corresponds respectively to many images or parametric texture, to reach the effect of image recognition.
Can be further understood by the following detailed description and accompanying drawings about the advantages and spirit of the present invention.
Description of drawings
Fig. 1 shows the process flow diagram according to the image-recognizing method of first specific embodiment of the present invention.
Fig. 2 further shows the possible detailed embodiment process flow diagram of step S13 among Fig. 1.
Fig. 3 shows the functional block diagram according to the pattern recognition device of second specific embodiment of the present invention.
Fig. 4 shows the functional block diagram that pattern recognition device shown in Figure 3 further comprises a selection module.
The reference numeral explanation
S11-S15: process step
10: pattern recognition device 11: judge module
12:DCT/IDCT module 13: comparison module
14: select module.
Embodiment
The characteristic that the present invention mainly can be quantized by the texture of digital picture is carried out discrete cosine transform at digital picture, to extract its textural characteristics and to carry out image recognition according to this textural characteristics.Discrete cosine transform mainly is to be applied to the JPEG compress technique at present, the image (for example BMP image) of other kind is transformed to the step that all can comprise discrete cosine transform in the program of jpeg image.
For example, be in the program of a jpeg image with a colored BMP image transformation, comprised steps such as grey scale transformation sampling, discrete cosine transform, quantification and coding.Owing to wherein comprised the step of discrete cosine transform, so this program will produce one group of discrete cosine transform coefficient corresponding to this image originally.This group discrete cosine transform coefficient can be in order to represent this BMP image texture features.On the other hand, carry out inverse discrete cosine transform at certain jpeg image and also can produce one group of discrete cosine transform coefficient corresponding to this jpeg image.
First specific embodiment according to the present invention is a kind of image-recognizing method.This method can be applicable to the identification of digital picture.See also Fig. 1, Fig. 1 shows the process flow diagram of this image-recognizing method.As shown in Figure 1, in this embodiment, this method is execution in step S11 at first, judges whether to exist one group of parametric texture corresponding to an image.
If the judged result of step S11 is that this method execution in step S12 does not judge whether to exist one group of discrete cosine transform coefficient corresponding to this image.If the judged result of step S12 is that this method execution in step S13 does not optionally carry out a discrete cosine transform or an inverse discrete cosine transform at this image, to produce this group discrete cosine transform coefficient.Then, this method execution in step S14 produces this group parametric texture according to this group discrete cosine transform coefficient.At last, this method execution in step S15 relatively should organize parametric texture and one group of target texture parameter, to produce a recognition result.If the judged result of step S11 is for being that then the direct execution in step S15 of this method relatively should organize parametric texture and one group of target texture parameter, to produce a recognition result.
If the judged result of step S12 is for being that then the direct execution in step S14 of this method produces this group parametric texture according to this group discrete cosine transform coefficient.Then, this method execution in step S15 relatively should organize parametric texture and one group of target texture parameter, to produce a recognition result.If this group parametric texture is corresponding to a target image, this recognition result can be represented the similarity degree of this image and this target image.
According to the present invention, for identification demand in the future, each BMP image is transformed to the discrete cosine transform coefficient that produces in the process of jpeg image and can be stored in advance.On the other hand, if before produced corresponding to the parametric texture of a certain image, its parametric texture also can be stored.By this, when the user required to discern this image once more, the method according to this invention and device promptly needn't recomputate the discrete cosine transform coefficient and the parametric texture of this image.
See also Fig. 2, Fig. 2 further shows the possible detailed embodiment of step S13 among Fig. 1.As shown in Figure 2, when the form of this image not simultaneously, the performed step of this method is also inequality.This embodiment is to be example with JPEG and two kinds of picture formats of BMP.In actual applications, the method certainly is applied to the identification of the image of other different-format.
Step S11 among Fig. 2, S12, S14 and the S15 all step with shown in Figure 1 are identical, therefore repeat no more.In this embodiment, if the judged result of step S12 is not for, this method execution in step S13A then, the form of judging this image is BMP form or jpeg format.If the form of this image is the BMP form, then this method execution in step S13B carries out this discrete cosine transform (DCT) at this image, to produce this group discrete cosine transform coefficient.In addition, before carrying out this discrete cosine transform, this method can be carried out grey scale transformation at this BMP image.On the other hand, if the form of this image is a jpeg format, then this method execution in step S13C carries out this inverse discrete cosine transform (IDCT) at this image, to produce this group discrete cosine transform coefficient.Behind step S13B and S13C, this method can further be stored this group discrete cosine transform coefficient.This method also can further be stored this group parametric texture after step S14.
In actual applications, based on the characteristic of discrete cosine transform coefficient, this group parametric texture can comprise one first smooth grain energy (E
DC1), first vertical texture/horizontal texture energy is than [(E
V1/ E
H1)] and one first oblique texture energy (E
S1).For example, very high if certain opens the pairing smooth grain energy of image, may represent that this image comprises a large amount of smooth regions.Accordingly, this group target texture parameter then can comprise a target smooth grain energy (E
DC), target vertical texture/horizontal texture energy is than (E
V/ E
H) and the oblique texture energy (E of a target
S).Because above-mentioned energy or energy ratio can be corresponding to this image texture features, therefore when this method execution in step S15, can be according to above-mentioned E
DC1, (E
V1/ E
H1), E
S1, E
DC, (E
V/ E
H) and E
STo produce this recognition result.Suppose that D represents this recognition result, then D can be expressed as:
Wherein, a, b and c are weighting coefficient.
In fact, user or computer system may be to wish by identifying in many candidate images and the immediate image of this target image.The method according to this invention also can further comprise the situation of considering a plural number candidate image.Suppose always to have N and open candidate image, wherein, N is a positive integer.The method according to this invention can be opened image at this N and be carried out as shown in Figure 1 step, produces the candidate's recognition result corresponding to each candidate image respectively.According to described candidate's recognition result, the method according to this invention can be opened by this N and be selected a result images the most similar to this target image in the candidate image.
In other words, described candidate's recognition result by this method gained, can learn that similarity degree that this N opens candidate image and this target image why, and this N can be opened candidate image and arrange, look user's demand and select the image the most similar or the image of other similarity degree to this target image according to the difference of similarity degree.
Therefore for example, suppose always to have 5 candidate images, carry out to obtain 5 candidate's recognition results after as shown in Figure 1 the step at every candidate image.If represented this candidate image of these 5 candidate's recognition results and the similarity degree of this target image are respectively 65%, 77%, 86%, 93% and 98%, then reach this image of 98% and be this result images of being asked with the similarity degree of this target image.
Second specific embodiment according to the present invention is a kind of pattern recognition device.See also Fig. 3, Fig. 3 shows the functional block diagram according to the pattern recognition device of second specific embodiment of the present invention.As shown in Figure 3, pattern recognition device 10 comprises a judge module 11, a DCT/IDCT module 12 and a comparison module 13.Judge module 11 is used to judge whether to exist corresponding to one group of discrete cosine transform coefficient of an image and/or corresponding to one group of parametric texture of this group discrete cosine transform coefficient.DCT/IDCT module 12 is electrically connected to judge module 11, if the judged result of judge module 11 is for denying, DCT/IDCT module 12 is promptly according to a form of this image, optionally carry out a discrete cosine transform (DCT) or an inverse discrete cosine transform (IDCT) at this image, producing this group discrete cosine transform coefficient, and produce this group parametric texture according to this group discrete cosine transform coefficient.Comparison module 13 is electrically connected to DCT/IDCT module 12, and is used for relatively this group discrete cosine transform coefficient and one group of target discrete cosine transform coefficient, to produce a recognition result.
In addition, if the judged result of judge module 11 exists for this group parametric texture, comparison module 13 promptly relatively should be organized the target texture parameter with this by the group parametric texture, to produce this recognition result.If the judged result of judge module 11 exists for this group discrete cosine transform coefficient, but this group parametric texture does not exist, DCT/IDCT module 12 promptly produces this group parametric texture according to this group discrete cosine transform coefficient, and comparison module 13 relatively should be organized the target texture parameter with this by the group parametric texture, to produce this recognition result.
For example, if this form is a BMP form, DCT/IDCT module 12 can be carried out this discrete cosine transform at this image.This device also can further comprise a grey scale transformation module (not showing in the drawings), before carrying out this discrete cosine transform in DCT/IDCT module 12, this image is carried out a grey scale transformation.On the other hand, if this form is a jpeg format, DCT/IDCT module 12 can be carried out this inverse discrete cosine transform at this image.This device can further comprise a memory module (not showing in the drawings).This memory module is electrically connected to DCT/IDCT module 12, and is used to store this group discrete cosine transform coefficient and/or this group parametric texture.
In fact, user or computer system may be to wish by identifying in many candidate images and the immediate image of this target image.Also can further contain the situation of considering a plural number candidate image according to device of the present invention.Suppose that total N opens candidate image, N is a positive integer.Judge module 11, DCT/IDCT module 12, comparison module 13 can be opened image at this N and produce candidate's recognition result corresponding to each candidate image respectively.As shown in Figure 4, pattern recognition device 10 can further comprise a selection module 14.Select module 14 to be electrically connected to comparison module 13, and according to described candidate's recognition result, open by this N and select a result images the most similar in the candidate image to this target image.
Compared to prior art, image-recognizing method of the present invention and device thereof are to utilize a discrete cosine transform or an anti-phase discrete cosine transform to obtain one group of discrete cosine transform coefficient corresponding to an image.Therefore because this group discrete cosine transform coefficient is corresponding to image texture features, can and/or reach the effect of image recognition corresponding to the parametric texture of discrete cosine transform coefficient by the many groups discrete cosine transform coefficient that relatively corresponds respectively to many images.Compared with traditional image recognition technology at carrying out calculation process between image and image and utilizing pixel that the comparison of pixel is come recognition image, not only accelerate the speed of image recognition according to image-recognizing method of the present invention and device thereof, also significantly saved the cost of image recognition.
By the above detailed description of preferred embodiments, be to wish to know more to describe feature of the present invention and spirit, and be not to come category of the present invention is limited with above-mentioned disclosed preferred embodiment.On the contrary, its objective is that hope can contain in the category of claim of being arranged in of various changes and tool equality institute of the present invention desire application.Therefore, the category of the claim that the present invention applied for should be done the broadest explanation according to above-mentioned explanation, contains the arrangement of all possible change and tool equality to cause it.
Claims (22)
1. image-recognizing method comprises the following step:
A) judge whether to exist corresponding to one group of discrete cosine transform coefficient of an image and/or corresponding to one group of parametric texture of this group discrete cosine transform coefficient;
B) if the judged result of step a) is not, form according to this image, optionally carry out a discrete cosine transform or an inverse discrete cosine transform, producing this group discrete cosine transform coefficient, and produce this group parametric texture according to this group discrete cosine transform coefficient at this image; And
C) relatively should organize parametric texture and one group of target texture parameter, to produce a recognition result.
2. the method for claim 1 further comprises the following step:
D), relatively should organize the target texture parameter with this by the group parametric texture, to produce this recognition result if the judged result of step a) exists for this group parametric texture.
3. the method for claim 1 further comprises the following step:
E) if the judged result of step a) exists for this group discrete cosine transform coefficient, but this group parametric texture does not exist, promptly produce this group parametric texture, and relatively should organize the target texture parameter with this by the group parametric texture, to produce this recognition result according to this group discrete cosine transform coefficient.
4. the method for claim 1, wherein if this form is a BMP form, step b) is to carry out this discrete cosine transform at this image.
5. method as claimed in claim 4 further comprises the following step:
F) before carrying out this discrete cosine transform, carry out a grey scale transformation at this image.
6. the method for claim 1, wherein if this form is a jpeg format, step b) is to carry out this inverse discrete cosine transform at this image.
7. the method for claim 1, wherein this group parametric texture comprises one first smooth grain ENERGY E
DC1, first vertical texture/horizontal texture energy compares E
V1/ E
H1And one first oblique texture energy E
S1, this group target texture parameter comprises a target smooth grain ENERGY E
DC, target vertical texture/horizontal texture energy compares E
V/ E
HAnd the oblique texture energy E of a target
S, and step c) is according to E
DC1, E
V1/ E
H1, E
S1, E
DC, E
V/ E
HAnd E
SProduce this recognition result.
8. method as claimed in claim 7, wherein, this recognition result D) can be represented as:
Wherein, a, b and c are weighting coefficient.
9. the method for claim 1, wherein should organize the target texture parameter corresponding to a target image, and a similarity degree of this image of this recognition result and this target image.
10. method as claimed in claim 9, wherein, N opens candidate image and comprises this image, and N is a positive integer, and this N opens in the candidate image each and open candidate image and correspond respectively to candidate's recognition result, and this method further comprises the following step:
G), open by this N and to select a result images the most similar in the candidate image to this target image according to described candidate's recognition result.
11. the method for claim 1 further comprises the following step:
H) after step b), store this group discrete cosine transform coefficient and/or this group parametric texture.
12. a pattern recognition device comprises:
One judge module is in order to judge whether to exist corresponding to one group of discrete cosine transform coefficient of an image and/or corresponding to one group of parametric texture of this group discrete cosine transform coefficient;
One DCT/IDCT module, be electrically connected to this judge module, if the judged result of this judge module is for denying, this DCT/IDCT module is according to a form of this image, optionally carry out a discrete cosine transform or an inverse discrete cosine transform at this image, producing this group discrete cosine transform coefficient, and produce this group parametric texture according to this group discrete cosine transform coefficient; And
One comparison module, this comparison module are electrically connected to this DCT/IDCT module, and are used for relatively this group discrete cosine transform coefficient and one group of target discrete cosine transform coefficient, to produce a recognition result.
13. device as claimed in claim 12, wherein, if the judged result of this judge module exists for this group parametric texture, this comparison module relatively should be organized the target texture parameter with this by the group parametric texture, to produce this recognition result.
14. device as claimed in claim 12, wherein, if the judged result of this judge module exists for this group discrete cosine transform coefficient, but this group parametric texture does not exist, this DCT/IDCT module promptly produces this group parametric texture according to this group discrete cosine transform coefficient, and this comparison module relatively should be organized the target texture parameter with this by the group parametric texture, to produce this recognition result.
15. device as claimed in claim 12, wherein, if this form is a BMP form, this DCT/IDCT module is to carry out this discrete cosine transform at this image.
16. device as claimed in claim 14 further comprises:
One grey scale transformation module, this grey scale transformation module is electrically connected to this DCT/IDCT module, and is used for before this DCT/IDCT module is carried out this discrete cosine transform this image being carried out a grey scale transformation.
17. device as claimed in claim 12, wherein, if this form is a jpeg format, this DCT/IDCT module is carried out this inverse discrete cosine transform at this image.
18. device as claimed in claim 12, wherein, this group parametric texture comprises one first smooth grain ENERGY E
DC1, first vertical texture/horizontal texture energy compares E
V1/ E
H1And one first oblique texture energy E
S1, this group target texture parameter comprises a target smooth grain ENERGY E
DC, target vertical texture/horizontal texture energy compares E
V/ E
HAnd the oblique texture energy E of a target
S, and this comparison module is according to E
DC1, E
V1/ E
H1, E
S1, E
DC, E
V/ E
HAnd E
SProduce this recognition result.
19. device as claimed in claim 18, wherein, this recognition result (D) can be represented as:
Wherein, a, b and c are weighting coefficient.
20. device as claimed in claim 12, wherein, this group target texture parameter is corresponding to a target image, and this recognition result is a similarity degree of this image and this target image.
21. device as claimed in claim 20, wherein, N opens candidate image and comprises this image, and N is a positive integer, and this N opens in the candidate image each and open candidate image and correspond respectively to candidate's recognition result, and this device further comprises:
One selects module, is electrically connected to this comparison module, and according to described candidate's recognition result, is opened by this N and selects a result images the most similar to this target image in the candidate image.
22. device as claimed in claim 12 further comprises:
One memory module, this memory module are electrically connected to this DCT/IDCT module, and are used to store this group discrete cosine transform coefficient and/or this group parametric texture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2007101382836A CN101359364B (en) | 2007-08-03 | 2007-08-03 | Image recognition method and image recognition device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2007101382836A CN101359364B (en) | 2007-08-03 | 2007-08-03 | Image recognition method and image recognition device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101359364A true CN101359364A (en) | 2009-02-04 |
| CN101359364B CN101359364B (en) | 2011-08-31 |
Family
ID=40331810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2007101382836A Expired - Fee Related CN101359364B (en) | 2007-08-03 | 2007-08-03 | Image recognition method and image recognition device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101359364B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102136074A (en) * | 2011-03-03 | 2011-07-27 | 浙江农林大学 | Man-machine interface (MMI) based wood image texture analyzing and identifying method |
| CN116383795A (en) * | 2023-06-01 | 2023-07-04 | 杭州海康威视数字技术股份有限公司 | Biological feature recognition method and device and electronic equipment |
-
2007
- 2007-08-03 CN CN2007101382836A patent/CN101359364B/en not_active Expired - Fee Related
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102136074A (en) * | 2011-03-03 | 2011-07-27 | 浙江农林大学 | Man-machine interface (MMI) based wood image texture analyzing and identifying method |
| CN102136074B (en) * | 2011-03-03 | 2012-09-26 | 浙江农林大学 | Man-machine interface (MMI) based wood image texture analyzing and identifying method |
| CN116383795A (en) * | 2023-06-01 | 2023-07-04 | 杭州海康威视数字技术股份有限公司 | Biological feature recognition method and device and electronic equipment |
| CN116383795B (en) * | 2023-06-01 | 2023-08-25 | 杭州海康威视数字技术股份有限公司 | Biological feature recognition method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101359364B (en) | 2011-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | An effective method for detecting double JPEG compression with the same quantization matrix | |
| JP2007104678A (en) | Method of classifying images in connection with jpeg compression history | |
| Wei et al. | Distortion function based on residual blocks for JPEG steganography | |
| CN112650875A (en) | House image verification method and device, computer equipment and storage medium | |
| Singh et al. | A self recoverable dual watermarking scheme for copyright protection and integrity verification | |
| EP1840836A1 (en) | Method of performing fast compression and decompression for image | |
| CN111753673A (en) | Video data detection method and device | |
| Fadl et al. | A proposed accelerated image copy-move forgery detection | |
| CN101359364B (en) | Image recognition method and image recognition device | |
| CN106709915B (en) | Image resampling operation detection method | |
| US8064707B2 (en) | Image recognition method and image recognition apparatus | |
| Shreelekshmi et al. | Undetectable least significant bit replacement steganography | |
| CN106055632B (en) | Video authentication method based on scene frame fingerprint | |
| CN100535926C (en) | Data processing method and apparatus, image processing method and apparatus, image sorting method and apparatus, and storage medium | |
| Yang et al. | Detecting doubly compressed JPEG images by factor histogram | |
| Xu et al. | Detecting image splicing using merged features in chroma space | |
| CN106845540B (en) | Image resampling operation interpolation type identification method | |
| CN1804869A (en) | Grayscale character dictionary generation apparatus | |
| CN113298112B (en) | Integrated data intelligent labeling method and system | |
| Meng et al. | Tamper detection for shifted double jpeg compression | |
| CN113298102B (en) | Training method and device for target classification model | |
| CN109118424B (en) | Image processing memory management method and management system for target detection | |
| CN114241253A (en) | Model training method, system, server and storage medium for illegal content identification | |
| CN112036501A (en) | Image similarity detection method based on convolutional neural network and related equipment thereof | |
| Fadl et al. | Fan search for image copy-move forgery detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110831 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |