CN100420269C - Picture reinforcing treatment system and treatment method - Google Patents

Abstract

The system includes following parts: image buffer; image analyzer in multiple scales connected to the image buffer; fuzzy logic controller (FLC) connected to the image analyzer in multiple scales; edge enhancement processor connected to image buffer and FLC; noise remover connected to connected to image buffer and FLC; synthesizer connected to image buffer, noise remover, edge enhancement processor, and FLC. The method includes following steps: based on fetched original image data to obtain edge information of original image; based on edge information of original image to obtain variance of gain and noise of original image; based on calculated data of edge enhanced image and noise removed image to obtain final processed data. Using edge analysis in multiple scales, under control of fuzzy logic, the invention realizes self-adaptive noise reduction process and edge enhancement process for images.

Description

A kind of image enhancement processing system and processing method

Technical field

The present invention relates to a kind of image processing system and processing method, especially a kind of image enhancement processing system and processing method of utilizing fuzzy logic adaptive control denoising and edge to strengthen.

Background technology

In vision signal is handled, need carry out denoising and edge enhancement process.Usually, utilize the low pass filter of certain frequency range to carry out noise reduction, in most of the cases, these frequency ranges still comprise useful information, therefore after denoising, can lose a lot of edge details information.Under the limited situation of image Compression and the bandwidth of video signal, also can lose image edge information.The sharpening of image border requires the loss with high pass filter compensating images edge.Noise reduction requires to reduce the radio-frequency component in the signal, and the edge strengthens the radio-frequency component that requires to increase in the signal.Therefore, noise reduction and edge sharpening are the processing of two contradictions.

Existing processing unit and processing method are to carry out rim detection and denoising on original image.On the result of denoising, strengthen image processing then, but, therefore can lose little details after the denoising through the denoising device because adopt the structure of serial with high passband ripple device.All be noise reduction and edge sharpening to be handled carry out serial process, that is: low-pass filtering vision signal noise reduction, high-pass filtering comes sharpen edges then.First noise reduction process loss marginal information and second sharpen edges process amplified noise, so the method can not make denoising and figure image intensifying reach optimum simultaneously.

Also have and utilize parallel organization to carry out that noise reduction and edge strengthen, but only the gentle vertical both direction of water strengthens the edge; Just handle through simple addition, fuzzy logic is not come the mutual of control module inside and intermodule flexibly.

Summary of the invention

The objective of the invention is deficiency, a kind of image processing system and processing method are provided, utilize the good noise reduction process of the next adaptive realization of fuzzy logic and realize that good edge strengthens at conventional images enhancement process system and processing method.

For achieving the above object, the invention provides a kind of image enhancement processing system, comprising:

One image buffer;

One multi-scale image analyzer is connected with described image buffer, is used to extract the marginal information of original image;

One fuzzy logic controller is connected with described multi-scale image analyzer, is used for according to the gain of the marginal information computed image of original image and the variance of calculating noise;

One edge enhancement process device is connected with fuzzy logic controller with described image buffer, is used for according to the signal after the gain calculating output edge enhancing of image;

One denoising device, it is connected with described image buffer and fuzzy logic control, is used for according to the signal after the variance calculating denoising of noise;

One synthesizer is connected with fuzzy logic controller with described image buffer, denoising device, edge enhancement process device, is used for drawing last output signal according to the signal after the enhancing of described edge, signal and the picture signal after the denoising;

This fuzzy logic controller comprises:

One edge classification of type module is connected with described multi-scale image analyzer;

One edge gain control module is connected with edge enhancement process device with described edge type sort module;

One noise average and variance statistical module are connected with the denoising device with described multi-scale image analyzer.

Above-mentioned multi-scale image analyzer comprises:

One gaussian filtering module is connected with described image buffer;

One first level and vertical Suo Beier filtration module are connected with described gaussian filtering module;

One first amplitude detection module is connected with fuzzy logic controller with vertical Suo Beier filtration module with described first level;

One second level and vertical Suo Beier filtration module are connected with described image buffer;

One direction detection module is connected with fuzzy logic controller with vertical Suo Beier filtration module with described second level;

One second amplitude detection module is connected with fuzzy logic controller with vertical Suo Beier filtration module with described second level;

One level and vertical high-pass filtering module are connected with described image buffer;

One the 3rd amplitude detection module is connected with fuzzy logic controller with vertical high-pass filtering module with described level.

In addition, above-mentioned edge enhancement process device can comprise:

One high-pass filtering module is connected with described image buffer;

One first multiplier is connected with the edge gain control module with described high-pass filtering module;

Bandpass filtering modules block in one is connected with described image buffer;

One second multiplier is connected with the edge gain control module with described middle bandpass filtering modules block;

One low bandpass filtering module is connected with described image buffer;

One the 3rd multiplier is connected with described low band pass filter module edge gain control module;

One first adder is connected with second multiplier with described first multiplier;

One second adder is connected with first adder with described the 3rd multiplier.

The present invention also provides a kind of image enhancement processing method, may further comprise the steps:

Step 1, multi-scale image analyzer read raw image data from image buffer, draw the marginal information of original image, in the input fuzzy logic controller;

Step 2, described fuzzy logic controller be according to the marginal information of original image, and the gain that draws original image in the input edge enhancement process device, and draws the variance of noise, in the input denoising device;

Step 3, edge enhancement process device draw the view data that strengthens through the edge according to the gain of original image and the raw image data that reads from image buffer, in the input synthesizer; The denoising device is according to the variance of noise and the raw image data that reads from image buffer, draws the view data after the denoising;

Step 4, synthesizer draw last deal with data according to described view data through the edge enhancing, view data and the raw image data after the denoising;

Above-mentioned steps 2 is specially:

Edge type sort module in step 21, the described fuzzy logic controller obtains the probability on the big limit of each pixel, middle limit and little limit according to the amplitude on the amplitude on described big limit, middle limit and the amplitude on little limit, in the input edge gain control module; Noise average in the described fuzzy logic controller and variance statistical module obtain the variance and the threshold value of noise according to the direction on described limit, are input in the denoising device;

Step 22, described edge gain control module obtain the gain on the gain on the big limit of original image, middle limit and the gain on little limit, and are input in the described edge enhancement process device according to the probability on described big limit, middle limit and little limit.

Above-mentioned steps 1 can be specially: raw image data is through the process of convolution of gaussian filtering module, obtain horizontal gradient and vertical gradient through first level and the process of convolution that vertical Suo Beier leads the mode piece then, the first amplitude detection module calculate horizontal gradient and vertical gradient mould and, i.e. the amplitude on big limit; Raw image data obtains horizontal gradient and vertical gradient through second level and the vertical process of convolution of Suo Beier filtration module, and the direction detection module obtains the direction on limit, the amplitude on limit during the second amplitude detection module obtains according to horizontal gradient and vertical gradient; Raw image data obtains horizontal gradient and vertical gradient through the process of convolution of level and vertical high-pass filtering module, and the 3rd amplitude detection module obtains the amplitude on little limit, and all is input in the fuzzy logic controller.

And the edge strengthens it according to the gain of original image and the raw image data that reads in the above-mentioned steps 3 from image buffer, draws the view data that strengthens through the edge and is specially:

After step 31, the processing of raw image data through the high-pass filtering module, with the gain on little limit mutually product obtain little limit deal with data, through in after the processing of bandpass filtering modules block, the deal with data on limit during product obtains mutually with the gain on middle limit, after the processing through low bandpass filtering module, with the gain on big limit mutually product obtain the deal with data on big limit;

The deal with data summation on the deal with data on step 32, described little limit deal with data, middle limit and big limit obtains the view data through the edge enhancing.

And the step 4 in each technical scheme can be specially:

Step 41, described synthesizer view data and the raw image data after according to described denoising draws noise data;

Step 42, the long-pending and noise data and the noise that calculate the described view data that strengthens through the edge and edge gain gain amass poor, sue for peace to the end deal with data with described original image again.

Therefore, the present invention utilizes multiple dimensioned edge analysis, has realized adaptive noise reduction process and edge of image enhancement process under the control of fuzzy logic.

Below by drawings and Examples, technical scheme of the present invention is described in further detail.

Description of drawings

Fig. 1 is the structural representation of image enhancement processing of the present invention system.

Fig. 2 is the multi-scale image analyzer in the image enhancement processing of the present invention system and the detailed structure schematic diagram of fuzzy logic controller.

Fig. 3 is the detailed structure schematic diagram of the edge enhancement process device in the image enhancement processing of the present invention system.

Fig. 4 is the level in the image enhancement processing of the present invention system and the vertical filter factor schematic diagram of Suo Beier filtration module and high-pass filtering module.

The schematic diagram that Fig. 5 divides for the edge direction of image enhancement processing of the present invention system.

Fig. 6 is the schematic diagram on the little limit of image enhancement processing of the present invention system, middle limit and big limit.

Fig. 7 is the schematic diagram of the different frequency bands of image enhancement processing of the present invention system.

Fig. 8 strengthens the function schematic diagram for the edge of image enhancement processing of the present invention system.

Fig. 9 is the schematic diagram of the window pixel of image enhancement processing of the present invention system.

Figure 10 is the low bandpass filtering module of image enhancement processing of the present invention system, the spectral characteristic schematic diagram of middle bandpass filtering modules block and high-pass filtering module.

Figure 11 is the denoising schematic diagram of image enhancement processing of the present invention system.

Figure 12 is the flow chart of image enhancement processing method of the present invention.

Embodiment

The present invention utilizes multiple dimensioned edge analysis, carries out adaptive noise reduction process and edge of image enhancement process under the control of fuzzy logic.

As shown in Figure 1, structural representation for image enhancement processing of the present invention system, comprise image buffer 1, the denoising device 2 that is connected with image buffer 1, edge enhancement process device 3 and multi-scale image analyzer 4, also comprise fuzzy logic controller 5, be connected with multi-scale image analyzer 4 with denoising device 2, edge enhancement process device 3, and synthesizer 6, be connected with fuzzy logic controller 5 with image buffer 1, denoising device 2, edge enhancement process device 3.

Each apparatus function is as follows:

Image buffer 1 is used to store raw image data;

Denoising device 2 is used for according to the signal after the variance calculating denoising of noise;

Edge enhancement process device 3 is used for according to the signal after the gain calculating output edge enhancing of image;

Multi-scale image analyzer 4 is used to extract the marginal information of original image;

Fuzzy logic controller 5 is used for according to the gain of the marginal information computed image of original image and the variance of calculating noise;

Synthesizer 6 is used for drawing last output signal according to the signal after the enhancing of described edge, signal and the picture signal after the denoising.

As shown in Figure 2, be the multi-scale image analyzer in the image enhancement processing of the present invention system and the detailed structure schematic diagram of fuzzy logic controller, multi-scale image analyzer 4 comprises: gaussian filtering module 40, be connected with image buffer 1, first level and vertical Suo Beier filtration module 41, be connected with gaussian filtering module 40, the first amplitude detection module 42 is connected with edge type sort module 50 in the fuzzy logic controller 5 with first level and vertical Suo Beier filtration module 41; Second level and vertical Suo Beier filtration module 43, be connected with image buffer 1, direction detection module 44, with second level and vertically 43 of Suo Beier filtering moulds be connected with variance statistical module 52 with noise average in the fuzzy logic controller 5, the second amplitude detection module 45 is connected with edge type sort module 50 in the fuzzy logic controller 5 with second level and vertical Suo Beier filtration module 43; Level and vertical high-pass filtering module 46 are connected with image buffer 1, and the 3rd amplitude detection module 47 is connected with edge type sort module 50 in the fuzzy logic controller 5 with described level and vertical high-pass filtering module 43.Fuzzy logic controller 5 comprises, edge type sort module 50, edge gain control module 51 are connected with edge enhancement process device 3 with edge type sort module 50; Noise average and variance statistical module 52 are connected with denoising device 2 with direction detection module 44 in the multi-scale image analyzer 4.

As shown in Figure 3, detailed structure schematic diagram for the edge enhancement process device in the image enhancement processing of the present invention system, comprise high-pass filtering module 30, be connected with image buffer 1, first multiplier 31 is connected with edge gain control module 51 in the fuzzy logic controller 5 with high-pass filtering module 30, middle bandpass filtering modules block 32, be connected with image buffer 1, second multiplier 33 is connected with edge gain control module 51 with middle bandpass filtering modules block 32; Low bandpass filtering module 34 is connected with image buffer 1; The 3rd multiplier 35 is connected with edge gain control module 51 with low bandpass filtering module 34; First adder 36 is connected with second multiplier 33 with first multiplier 31; Second adder 37 is connected with first adder 36 with the 3rd multiplier 35.

Multi-scale image analyzer 4 carries out preliminary treatment to original image and obtains various information, comprising: the position on limit, direction, power, noise level, composition of picture material or the like.Note: Iorg is an original image information, Gau leads mode piece 40 for Gauss, SH and SV are the Suo Beier filtration module 41 of first level and vertical direction, Hf and Vf are the high-pass filtering module 46 of level and vertical direction, it at first is original image and gaussian filtering module 40 convolution that the first via is treated to, then with first level and vertically Suo Beier lead mode piece 41 and carry out convolution and obtain the gradient G h of level and vertical gradient G v, computing formula is as follows:

$\mathrm{Gh}=\mathrm{SH}\⊗(\mathrm{Iorg}\⊗\mathrm{Gau})$ With $\mathrm{Gv}=\mathrm{SV}\⊗(\mathrm{Iorg}\⊗\mathrm{Gau}),$

Be convolution operator, as shown in Figure 4 the filter coefficient schematic diagram of level and vertical Suo Beier filtration module and high-pass filtering module.

By the first amplitude detection module, 42 calculated level gradients and vertical gradient mould and, i.e. the amplitude El on big limit (LargeEdge): El=abs (Gh)+abs (Gv).The second the tunnel is that original image directly leads mode piece 43 with second level and vertical Suo Beier and carries out convolution, obtains horizontal gradient Gx and vertical gradient G y, and computing formula is as follows:

$\mathrm{Gx}=\mathrm{SH}\⊗\mathrm{Gau}$ With $\mathrm{Gy}=\mathrm{SV}\⊗\mathrm{Gau},$

Be convolution operator.

Direction detection module 44 is according to rule:

The direction Ed that obtains the limit determines four direction, as shown in Figure 5, the schematic diagram that edge direction is divided, the included scope of all directions is understood in the figure acceptance of the bid, the I zone is the vertical direction zone, the II zone is the horizontal direction zone, and the III zone is to the zone, angular direction, and the IV zone is for opposing the zone, angular direction.Not care about one's appearance value Em:Em=|Gx|+|Gy| during the second amplitude detection module 45 obtains simultaneously,

Third Road is that raw image data and level and vertical high-pass filtering module 46 are carried out convolution, obtains horizontal gradient Hx and vertical gradient Hy:

$\mathrm{Hx}=\mathrm{Hf}\⊗\mathrm{Gau}$ With $\mathrm{Hy}=\mathrm{Vf}\⊗\mathrm{Gau},$

Be convolution operator.

The two mould and as little not care about one's appearance value Es, that is: Es=abs (Hx)+abs (Hy).

Edge type sort module 50 is divided three classes the type on limit according to the acutance of edge transition: little limit, middle limit and big limit.Be illustrated in figure 6 as the schematic diagram on little limit, middle limit and big limit, what portrayed on little limit is the very big limit of acutance, and what portrayed on middle limit is the slightly slow limit of acutance, and what portrayed on big limit is very slow limit.The Es that multi-scale image analyzer 4 is sent here, three amounts itself of Em and El are exactly the component of each pixel of image on three yardsticks, edge type sort module 50 the value of these three amounts as this point belong to little, in, the probability on big limit.

As shown in Figure 7, be the schematic diagram of different frequency bands, A is big sideband, and B is middle sideband, and C is little sideband, Es, and three amounts of Em and El belong to different frequency bands respectively.For each pixel, Es, Em and El represent this high, in and the shared proportion of low-frequency band.The edge that is illustrated in figure 8 as image enhancement processing of the present invention system strengthens the function schematic diagram, adjusts final gain with linear function and controls three high passband ripple devices.

For little limit

Sg＝Ka_s*Es+Kb_s；

If Sg＞Up_limit, Sg=Up_limit so;

If Sg＜0, Sg=0 so;

For middle limit

Mg＝Ka_m*Es+Kb_m；

If Mg＞Up_limit, Mg=Up_limit so;

If Mg＜0, Mg=0 so;

For big limit

Lg＝Ka_l*Es+Kb_l；

If Lg＞Up_limit, Lg=Up_limit so;

If Lg＜0, Lg=0 so;

Edge gain controller 51 is Es like this, and Em and El change into three gains Sg, Mg, Lg, and wherein, Ka_s, Ka_m, Ka_l, Kb_s, Kb_m, Kb_l, Up_limit are the adjustable parameters of user.Total gain divides two parts: self adaptation part and standing part (the manual part of user).Ka_s, Ka_m, Ka_l control the self adaptation part.Kb_s, Kb_m, Kb_l are standing parts.Can control total gain very flexibly like this.Reach the image effect that adjusts.

Along the direction on limit, the grey scale change of image is smaller, and these little variations are mainly caused by noise.Being illustrated in figure 9 as the schematic diagram of the window pixel of image enhancement processing of the present invention system, is example with the 3*3 window and with the directions on 45 degree limits, based on this supposition, and the statistical module 52 of noise average and variance, mean value is:

Nm＝{2*(P1+P3+P5+P7)+2*(P2+P6)+4*P4}/16，

Noise variance is:

Nv＝{abs(P1-P3)+abs(P5-P7)+abs(P4-P2)+abs(P4-P6)}/4

Total threshold value is: Nt=Nm+Ng*Nv

N _gBe programmable, can be used as register parameters.

Denoising device 3 carries out nonlinear filtering to be handled, and parameter N m and Nv according to the statistical module 52 of noise average in the fuzzy logic controller 5 and variance transmits have

The filtering result is:

${\mathrm{PN}}_{\mathrm{out}}=\frac{\underset{i=0}{\overset{s}{\mathrm{\Σ}}}(S_i*{\mathrm{\δ}}_i)}{\underset{i=0}{\overset{s}{\mathrm{\Σ}}}{\mathrm{\δ}}_i}$

The average Nm of nonlinear filtering and variance Nv determine according to the direction on limit respectively.Gain is that the user adjusts according to the noise level of image.The direction of vertical edges, grey scale change is bigger.Along the direction on limit, gray scale has good consistency, and therefore the variance at this directional statistics is exactly the local variance of noise.Such nonlinear filtering wave energy well keeps the acutance on limit, also the denoising effect that can obtain.

Processing in the edge enhancement process device 3 is as follows, raw image data is through low bandpass filtering module (LBPF), the value that middle bandpass filtering modules block (BPF) and high-pass filtering module (HPF) (the frequency response schematic diagram of three two-dimensional filtering modules as shown in figure 10) obtain is designated as respectively: LBP, BP and HP, original image is designated as: Iorg

$\mathrm{LBP}=\mathrm{Iorg}\⊗\mathrm{LBPF};$

$\mathrm{BP}=\mathrm{Iorg}\⊗\mathrm{LBP};$

$\mathrm{HP}=\mathrm{Iorg}\⊗\mathrm{HP};$

This three part has been delineated image frequency content from low to high respectively.For each pixel, Lg, Mg and Sg are illustrated in this three kinds of proportions that composition is shared.Can adjust picture material low, that neutralization is high according to image adaptive, also can adjust according to the interested content of user.

The output of three filtration modules of gain controlling that fuzzy logic controller transmits, the result is:

PE _out＝LBP*Lg+BP*Mg+HP*Sg

According to the contained different frequency composition in limit, three gain Lg of fuzzy logic controller, Mg and Sg control low bandpass filtering module, middle bandpass filtering modules block and high-pass filtering module respectively.PEout is final high frequency output.

Denoising device and edge intensifier all utilize the content of original image to obtain two parts result: denoising result PNout and image detail be PEout as a result.In the abundant zone of the details of image, it is more to wish to keep PEout.In level and smooth district, noise effect is apparent in view, wishes that denoising result PNout is more.

In texture area, direction is rambling.So the variance Nv that the statistical module 52 of noise average and variance is obtained is very big, should suppress denoising this moment.

The ratio TNg of denoising is:

(1) TNg=1, if Nv＜=NT, NT is that the user is according to the setting of image global noise.

(2) TNg=1-Ka_N* (NT-Nv), as TNg＜=KT, if TNg=KT so is Nv＞NT; KT is the adjustable parameter of user.

As shown in figure 11, be the denoising schematic diagram, the ratio TEg=1-TNg of the data after the output edge enhancement process, the visual pixel PNout after the adaptive threshold output denoising of denoising device 3 usefulness fuzzy logics; Then the noise of image is: (Iorg-PNout), this part is the radio-frequency component of image.The useful high-frequency information of edge enhancement process device 3 outputs is: PEout; Two parts radio-frequency component is controlled in two gains of fuzzy logic controller 5 usefulness (TEg and TNg), and these two gains are programmable.Synthesizer 6 utilizes: the output PNout of denoising device 2, gain TEg of noise average and 52 outputs of variance statistical disposition module in the output PEout of edge processor 3 and the fuzzy logic controller 5 (ratios of the data after the output edge enhancement process) and TNg (ratio of denoising); Finally be output as:

Pout＝TEg*PEout-TNg*(Iorg-PNout)+Iorg。

As shown in figure 12, be the flow chart of image enhancement processing method of the present invention.Concrete steps are as follows:

Step 101, multi-scale image analyzer read raw image data from image buffer, draw the marginal information of original image, in the input fuzzy logic controller;

In this step, detailed processing can for:

Raw image data is through the process of convolution of gaussian filtering module, obtain horizontal gradient and vertical gradient through first level and the process of convolution that vertical Suo Beier leads the mode piece then, the first amplitude detection module calculate horizontal gradient and vertical gradient mould and, i.e. the amplitude E1 on big limit; Raw image data is through the process of convolution of second level and vertical Suo Beier filtration module, obtain horizontal gradient Gx and vertical gradient Gy, the direction detection module obtains the direction Ed on limit according to horizontal gradient and vertical gradient, the amplitude Em on limit during the second amplitude detection module obtains; Raw image data obtains horizontal gradient Hx and vertical gradient Hy through the process of convolution of level and vertical high-pass filtering module, and the 3rd amplitude detection module obtains the amplitude Es on little limit, and all is input in the fuzzy logic controller;

Step 102, described fuzzy logic controller be according to the marginal information Es of original image, and Em and E l draw gain Sg, Mg and the Lg of original image, are input in the edge enhancement process device, and draw the variance Nv and the threshold value Nt of noise, in the input denoising device;

Processing method in this step is a lot, and present embodiment is specially:

Edge type sort module in step 1021, the described fuzzy logic controller is according to the amplitude Em on the amplitude El on described big limit, middle limit and the amplitude Es on little limit, obtain the probability Lg on the big limit of each pixel, middle limit and little limit, Mg and Sg are in the input edge gain control module; Noise average in the described fuzzy logic controller and variance statistical module obtain the variance Nv and the threshold value Nt of noise according to the direction on described limit, are input in the denoising device;

Step 1022, described edge gain control module are according to the probability Lg on described big limit, middle limit and little limit, and Mg and Sg obtain the gain Mg on the gain Lg on the big limit of original image, middle limit and the gain Sg on little limit, and are input in the described edge enhancement process device;

Step 103, edge enhancement process device are according to the gain Lg of original image, and Mg and Sg and the raw image data that reads from image buffer draw the view data PEout that strengthens through the edge, in the input synthesizer; The denoising device is according to the variance Nv of noise and the raw image data that reads from image buffer, draws the view data PNout after the denoising;

Edge booster in this step is according to the gain Lg of original image, Mg and Sg and the raw image data that from image buffer, reads, and it is also a lot of to draw the view data method that strengthens through the edge, is in detail in the present embodiment:

After step 1031, the processing of raw image data through the high-pass filtering module, with the gain on little limit mutually product obtain little limit deal with data, through in after the processing of bandpass filtering modules block, the deal with data on limit during product obtains mutually with the gain on middle limit, after the processing through low bandpass filtering module, with the gain on big limit mutually product obtain the deal with data on big limit;

The deal with data summation on the deal with data on step 1032, described little limit deal with data, middle limit and big limit obtains the view data PEout through the edge enhancing;

Step 104, synthesizer draw last deal with data according to described view data through the edge enhancing, view data and the raw image data after the denoising.

This step in this method is in detail:

Step 1041, described synthesizer view data and the raw image data after according to described denoising draws noise data (Iorg-PNout);

Step 1042, the described long-pending TEg*PEout that gains through the view data and the edge of edge enhancing of calculating, with the poor TEg*PEout-TNg* (Iorg-PNout) of the long-pending TNg* (Iorg-PNout) of noise data and noise gain, sue for peace to the end deal with data TEg*PEout-TNg* (Iorg-PNout)+Iorg with described original image again.

Therefore, the present invention utilizes multiple dimensioned edge analysis, has realized adaptive noise reduction process and edge of image enhancement process under the control of fuzzy logic.

It should be noted last that, above embodiment is only unrestricted in order to technical scheme of the present invention to be described, although the present invention is had been described in detail with reference to preferred embodiment, those of ordinary skill in the art is to be understood that, can make amendment or be equal to replacement technical scheme of the present invention, and not breaking away from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of the claim scope of the present invention.

Claims (8)

1. image enhancement processing system, comprising:

One image buffer;

One multi-scale image analyzer is connected with described image buffer, is used to extract the marginal information of original image;

One fuzzy logic controller is connected with described multi-scale image analyzer, is used for according to the gain of the marginal information computed image of original image and the variance of calculating noise;

One edge enhancement process device is connected with fuzzy logic controller with described image buffer, is used for according to the signal after the gain calculating output edge enhancing of image;

One denoising device, it is connected with described image buffer and fuzzy logic control, is used for according to the signal after the variance calculating denoising of noise;

One synthesizer is connected with fuzzy logic controller with described image buffer, denoising device, edge enhancement process device, is used for drawing last output signal according to the signal after the enhancing of described edge, signal and the picture signal after the denoising;

Described fuzzy logic controller comprises:

One edge classification of type module is connected with described multi-scale image analyzer;

One edge gain control module is connected with edge enhancement process device with described edge type sort module;

One noise average and variance statistical module are connected with the denoising device with described multi-scale image analyzer.

2. image enhancement processing according to claim 1 system, wherein said multi-scale image analyzer comprises:

One gaussian filtering module is connected with described image buffer;

One first level and vertical Suo Beier filtration module are connected with described gaussian filtering module;

One first amplitude detection module is connected with fuzzy logic controller with vertical Suo Beier filtration module with described first level;

One second level and vertical Suo Beier filtration module are connected with described image buffer;

One direction detection module is connected with fuzzy logic controller with vertical Suo Beier filtration module with described second level;

One second amplitude detection module is connected with fuzzy logic controller with vertical Suo Beier filtration module with described second level;

One level and vertical high-pass filtering module are connected with described image buffer;

One the 3rd amplitude detection module is connected with fuzzy logic controller with vertical high-pass filtering module with described level.

3. image enhancement processing according to claim 1 system, wherein said edge enhancement process device comprises:

One high-pass filtering module is connected with described image buffer;

One first multiplier is connected with the edge gain control module with described high-pass filtering module;

Bandpass filtering modules block in one is connected with described image buffer;

One second multiplier is connected with the edge gain control module with described middle bandpass filtering modules block;

One low bandpass filtering module is connected with described image buffer;

One the 3rd multiplier is connected with described low band pass filter module edge gain control module;

One first adder is connected with second multiplier with described first multiplier;

One second adder is connected with first adder with described the 3rd multiplier.

4. image enhancement processing method, comprising following steps:

Step 1, multi-scale image analyzer read raw image data from image buffer, draw the marginal information of original image, in the input fuzzy logic controller;

Step 2, described fuzzy logic controller be according to the marginal information of original image, and the gain that draws original image in the input edge enhancement process device, and draws the variance of noise, in the input denoising device;

Step 3, edge enhancement process device draw the view data that strengthens through the edge according to the gain of original image and the raw image data that reads from image buffer, in the input synthesizer; The denoising device is according to the variance of noise and the raw image data that reads from image buffer, draws the view data after the denoising;

Step 4, synthesizer draw last deal with data according to described view data through the edge enhancing, view data and the raw image data after the denoising;

Described step 2 is specially:

Edge type sort module in step 21, the described fuzzy logic controller obtains the probability on the big limit of each pixel, middle limit and little limit according to the amplitude on the amplitude on described big limit, middle limit and the amplitude on little limit, in the input edge gain control module; Noise average in the described fuzzy logic controller and variance statistical module obtain the variance and the threshold value of noise according to the direction on described limit, are input in the denoising device;

Step 22, described edge gain control module obtain the gain on the gain on the big limit of original image, middle limit and the gain on little limit, and are input in the described edge enhancement process device according to the probability on described big limit, middle limit and little limit.

5. image enhancement processing method according to claim 4, wherein said step 1 is specially: raw image data is through the process of convolution of gaussian filtering module, obtain horizontal gradient and vertical gradient through first level and the process of convolution that vertical Suo Beier leads the mode piece then, the first amplitude detection module calculate horizontal gradient and vertical gradient mould and, i.e. the amplitude on big limit; Raw image data obtains horizontal gradient and vertical gradient through second level and the vertical process of convolution of Suo Beier filtration module, and the direction detection module obtains the direction on limit, the amplitude on limit during the second amplitude detection module obtains according to horizontal gradient and vertical gradient; Raw image data obtains horizontal gradient and vertical gradient through the process of convolution of level and vertical high-pass filtering module, and the 3rd amplitude detection module obtains the amplitude on little limit, and all is input in the fuzzy logic controller.

6. according to claim 4 or 5 described image enhancement processing methods, the edge booster draws the view data that strengthens through the edge and is specially according to the gain of original image and the raw image data that reads from image buffer in the wherein said step 3:

After step 31, the processing of raw image data through the high-pass filtering module, with the gain on little limit mutually product obtain little limit deal with data, through in after the processing of bandpass filtering modules block, the deal with data on limit during product obtains mutually with the gain on middle limit, after the processing through low bandpass filtering module, with the gain on big limit mutually product obtain the deal with data on big limit;

The deal with data summation on the deal with data on step 32, described little limit deal with data, middle limit and big limit obtains the view data through the edge enhancing.

7. according to claim 4 or 5 described image enhancement processing methods, wherein said step 4 is specially:

Step 41, described synthesizer view data and the raw image data after according to described denoising draws noise data;

Step 42, the long-pending and noise data and the noise that calculate the described view data that strengthens through the edge and edge gain gain amass poor, sue for peace to the end deal with data with described original image again.

8. image enhancement processing method according to claim 6, wherein said step 4 is specially:

Step 41, described synthesizer view data and the raw image data after according to described denoising draws noise data;

Step 42, the long-pending and noise data and the noise that calculate the described view data that strengthens through the edge and edge gain gain amass poor, sue for peace to the end deal with data with described original image again.

Images