CN1208740C - Method for purifying noise and sharpening digital image - Google Patents

Abstract

The present invention relates to a method for simultaneously denoising and sharpening digital images, particularly to a method of synchronous processing for denoising and sharpening images. The method comprises the steps that firstly, a brightness diagram comprising brightness values of image pixels in digital images is worked out; secondly, an activity diagram is worked out from the brightness diagram, the activity value of each pixel in the diagram conforms to the activity surrounding the pixel; then, the activity magnitude of the pixel is compared with three predetermined threshold values, the activity diagram is modified by the three predetermined regulations, and the modified activity diagram is added back to the brightness channel of digital images. The three predetermined regulations respectively have a denoising function, a moderate sharpening function and a high sharpening function.

Description

The digital image noise purifies and sharpening method

Technical field

The present invention relates to a kind of image sharpening and noise purification method thereof, particularly, relate to a kind of method of image being carried out denoising and sharpening synchronous processing.

Background technology

The digital image treatment technology is that multipath improves picture quality and opened wide the gate, no matter be that its advantage all is obvious to all to motion or static image.

In the process that image is ingested, its pixel value always is subjected to the influence from various noise source random noises, in fact, every width of cloth image all comprises noise to a certain degree, and for example: the image that absorbs on traditional film comprises because the film particle that chemical reaction process produces; Equally, also comprise noise in the image that digital camera is directly caught, as shot noise, dark current noise and fixing photoreceptor noise or the like.

Noise effect picture quality, therefore suitably deletion or minimizing noise are necessary often.The natural property of light has determined to realize that a complete muting system is impossible, but can improve the ratio of signal and noise by certain technology, makes its high existence that can ignore noise to the observer relatively.

As everyone knows, strengthen the minimizing that not only can promote noise from the view data of optical detection device, also can make image sharpening simultaneously.Image sharpening has improved the acutance of digitized map corner edge, therefore can improve the quality of image usually.The U.S. 5,038, No. 388 patents have described the technology of image sharpening in detail.In many enhancing operations, denoising and image sharpening are two operations important and that adopt often.Improve acutance by suppressing noise, strengthen picture quality.But denoising and image sharpening interior last be contradiction, in fact all devices of catching image have the characteristic of low-pass filtering, just mean the high fdrequency component that has weakened image.Unfavorable low-pass filtering will reduce image definition.What make worse off is that the high fdrequency component that suppresses image all is inclined in nearly all noise purification run, thereby increases the loss of sharpness.Therefore, the also very necessary sharpening image manipulation that adopts.

As usual, a width of cloth digital image is successively carried out denoising and sharpening, denoising is prior to sharpening, because noise also can be exaggerated in sharpening process usually.Up to the present, noise removing and sharpening generally are separately designs.The main disadvantage of this method is that sharpening does not know that when operation noise purifies situation about having moved, thereby can't select best sharpening method.In addition, the parameter setting of selected sharpening method often is not best yet.

Therefore need provide a kind of method that can be applicable to digital image denoising and sharpening, the method is carried out to produce high-quality image denoising and sharpening synchronously under the situation of understanding operation each other mutually.

Summary of the invention

Problem to be solved by this invention just provides a kind of method, and this method logarithmic code image is simultaneously carried out denoising and sharpening operation.This method comprises following step:

At first, calculate luminance graph from digital image, it comprises the brightness value of image pixel;

Secondly, calculate activity diagram again from luminance graph, the activity value of each pixel among its figure meets the activity around this pixel.

Then, compare with the activity magnitude of this pixel and three threshold values that are predetermined, and compare and adjust, the luminance channel of corrected activity diagram add-back digital image.

If the activity magnitude of this pixel is the threshold value that is predetermined less than first, this activity then is judged as noise, with first predetermined adjustment it is revised earlier, is increased to the brightness value of this pixel then;

If the activity magnitude of this pixel is the threshold value that is predetermined greater than first, less than second predetermined threshold value, this activity then is judged as based between noise and the edge, will be left intact to this pixel;

If the activity magnitude of this pixel is the threshold value that is predetermined greater than second, less than the 3rd threshold value that is predetermined, it is weak edge that this activity then is judged as, and with second predetermined adjustment it is revised earlier, is increased to the brightness value of this pixel then;

If the activity magnitude of this pixel is greater than the 3rd threshold value that is predetermined, it is strong edge that this activity then is judged as, and with the 3rd predetermined adjustment it is revised earlier, is increased to the brightness value of this pixel then.

First predetermined adjustment is the denoising function.

Second predetermined adjustment is the sharpening function of appropriateness.

The 3rd predetermined adjustment is the sharpening function of height.The adjustment of all activity diagrams (AM) is all passed through lookup table (LUT) and realized, and is as follows:

$\mathrm{AMLUT}\left(x\right)=\left\{\begin{array}{cc}\mathrm{mx}& x\&le;-x_3\\ m{x}_3\&times;\left(\frac{x+x_2}{x_3-x_2}\right)& -x_3<x\&le;-x_2\\ 0& -x_2<x\&le;-x_1\\ \mathrm{Avg}-Y& -x_1<x\&le;x_1\\ 0& x_1<x\&le;x_2\\ m{x}_3\&times;\left(\frac{x-x_2}{x_3-x_2}\right)& x_2<x\&le;x_3\\ \mathrm{mx}& x_3<x\end{array}\right.$

Wherein, x is the activity magnitude of pixel, and x1, x2, x3 are respectively predetermined three threshold values.

The invention has the advantages that against existing technologies:

Method proposed by the invention logarithmic code image simultaneously purifies noise and sharpening operation.This denoising and sharpening are carried out method of operating in the mode of being mutually related, and can farthest improve the quality of image.In addition, purification and sharpening use look-up table also to guarantee the high-speed of processing.

Description of drawings

Figure one is the permutation table of 5 pixels * 5 pixels.

Figure two is picture specifications of the embodiment of the invention.

Figure three is picture specifications of the look-up table that is suitable for of the present invention.

Below be the Reference numeral of each several part among figure two and the figure three:

10 computer systems

20 main frames

30 displays

40 keyboards

50 mouses

52 cursors

55 CD drive (CD-ROM)

56 printers

57 (read-only) CD

61 disks

62 personal computer cards (PC card)

80LUT

Embodiment

At first, calculate luminance graph from digital image, it comprises the brightness value of image pixel;

Secondly, calculate activity diagram again from luminance graph, the activity value of each pixel among its figure meets the activity around this pixel.

Then, compare with the activity magnitude of this pixel and three threshold values that are predetermined, and compare and adjust, the luminance channel of corrected activity diagram add-back digital image.

If the activity magnitude of this pixel is the threshold value that is predetermined less than first, this activity then is judged as noise, with first predetermined adjustment it is revised earlier, is increased to the brightness value of this pixel then;

If the activity magnitude of this pixel is the threshold value that is predetermined greater than first, less than second predetermined threshold value, this activity then is judged as based between noise and the edge, will be left intact to this pixel;

If the activity magnitude of this pixel is the threshold value that is predetermined greater than second, less than the 3rd threshold value that is predetermined, it is weak edge that this activity then is judged as, and with second predetermined adjustment it is revised earlier, is increased to the brightness value of this pixel then;

If the activity magnitude of this pixel is greater than the 3rd threshold value that is predetermined, it is strong edge that this activity then is judged as, and with the 3rd predetermined adjustment it is revised earlier, is increased to the brightness value of this pixel then.

First predetermined adjustment is the denoising function.

Second predetermined adjustment is the sharpening function of appropriateness.

The 3rd predetermined adjustment is the sharpening function of height.

The monochromatic digital image of one width of cloth comprises a passage, just, and brightness or intensity passage.One width of cloth colored digital image is by redness, green, three blue passages form.Substantially, the edge of the acutance of image by the brightness of image passage reflects.Therefore, luminance channel is defined as the Y passage, will at first be derived.A formula of reasonably deriving the Y passage is as follows:

Y＝(red+2*green+blue)/4

In general, noise and edge all are considered as the endemic activity among a small circle of image.For instance, the pixel of the black matrix Y mark that is listed as at image m row and n shows in 5 pixels * 5 pixel coverages, sees accompanying drawing one.

For near the activity the pixel of weighing black matrix Y mark, following column matrix is used to the Y passage width of cloth activity diagram (AM) of deriving

$\left[\begin{array}{ccccc}-1& 0& -1& 0& -1\\ 0& 0& 0& 0& 0\\ -1& 0& 8& 0& -1\\ 0& 0& 0& 0& 0\\ -1& 0& -1& 0& -1\end{array}\right]\&times;\frac{1}{8}$

Promptly

$\mathrm{AM}=\frac{(8Y_{m,n}-Y_{m-2,n-2}-Y_{m-2,n}-Y_{m-2,n+2}-Y_{m,n-2}-Y_{m,n+2}-Y_{m+2,n-2}-Y_{m+2,n}-Y_{m+2,n+2})}{8}$

Activity diagram is further revised (AMLUT) by a look-up table.This look-up table shows in figure three graphically.

AMLUT is made up of following:

$\mathrm{AMLUT}\left(x\right)==\left\{\begin{array}{cc}{m}_{1}x& x\&le;-x_3\\ m_2{x}_3\&times;\left(\frac{x+x_2}{x_3-x_2}\right)& -x_3<x\&le;-x_2\\ 0& -x_2<x\&le;-x_1\\ \mathrm{Avg}-Y& -x_1<x\&le;x_1\\ 0& x_1<x\&le;x_2\\ m_2{x}_3\&times;\left(\frac{x-x_2}{x_3-x_2}\right)& x_2<x\&le;x_3\\ m_{1}x& x_3<x\end{array}\right.$

Wherein x is the activity magnitude of pixel, x ₁, x ₂, x ₃Be respectively predetermined threshold value, m ₁, m ₂It is given parameter.

When-x ₁＜x≤x ₁The time, x is considered to noise, is different from picture structure such as edge.In this case, following column operations will be used for calculating the Avg value.

At first, four direction mean value calculates according to following each row equation:

$\mathrm{AvgH}=\frac{(Y_{m,n}+Y_{m,n-2}+Y_{m,n-1}+Y_{m,n+1}+Y_{m,n+2})}{5};$

$\mathrm{AvgV}=\frac{Y_{m,n}+Y_{m-2,n}+Y_{m-1,n}+Y_{m+1,n}+Y_{m+2,n}}{5};$

$\mathrm{AvgL}=\frac{Y_{m,n}+Y_{m-2,n-2}+Y_{m-1,n-1}+Y_{m+1,n+1}+Y_{m+2,n+2}}{5};$

$\mathrm{AvgR}=\frac{Y_{m,n}+Y_{m-2,n+2}+Y_{m-1,n+1}+Y_{m+1,n-1}+Y_{m+2,n-2}}{5};$

Then, in the direction of level, definition DH _{M, n}=| Y _{M, n}-AvgH| and

DDH＝DH _m，n+DH _m，n-2+DH _m，n-1+DH _m，n+1+DH _m，n+2

Similarly, in vertical direction, diagonal and back-diagonal direction, define

DDV＝DV _m，n+DV _m-2，n+DV _m-1，n+DV _m+1，n+DV _m+2，n

DDL＝DL _m，n+DL _m-2，n-2+DL _m-1，n-1+DL _m+1，n+1+DL _m+2，n+2

DDR＝DR _m，n+DR _m+2，n-2+DR _m+1，n-1+DR _m-1，n+1+DR _m-2，n+2

If a pixel is on the edge, noise purifies and carries out along the edge only to guarantee that the edge is not darkened.For this reason, (DDH, DDV, DDL, DDR) middle minimum value need be identified, and Dui Ying directivity mean value is selected as Avg then, just,

Avg＝(AvgP|DDP＜＝DDQ，for?all?Q！＝P，Pin{H，V，L，R}，Qin{H，V，L，R})

At last, through corrected Y passage (purifying and sharpening) be through noise:

Y’＝Y+AMLUT(AM)

In above-mentioned invention, can carry out denoising and sharpening operation simultaneously, thereby make the sharpening operation know the operation of noise removing fully.

Know that in above description the present invention can preferentially be implemented in specific software program.But the expert of this area can be easy to realize this software equally in hardware.

Fig. 1 shows that is implemented a computer system 10 of the present invention.Computer system 10 comprises the main frame 20 based on microprocessor, is used for acceptance and process software programs and moves other processing capacity.A display 30 links to each other with main frame 20, shows the user data relevant with software.A keyboard 40 also links to each other with main frame 20, can make the user to the software input information.In addition, mouse 50 alternative keyboard 40 inputs, it can be used for the cursor 52 on the mobile display 30 and carries out option.

Compact disc read-only memory (read-only optical disc) 55 links to each other with main frame, accepts software program and Input Software program and the method for other information to main frame 20 are provided by read-only optical disc 57.Read-only optical disc 57 comprises a software program usually.In addition, disk 61 also comprises a software program, can insert main frame and be used for the Input Software program.Further, able to programme based on the main frame 20 of microprocessor, know in the technology, be used for the storage inside software program.A printer 56 links to each other with the main frame 20 that with the microprocessor is base, is used for the hard copy of the output of printing computer system 10.

Image can show on display 30 by a personal computer card (PC card).As what knew in the past, personal computer storage international association card (pcmcia card) electronic personal computer card 62 that comprises digitized image.Personal computer card 62 is inserted in the main frame 20 based on microprocessor at last, is used for display image on display 30.

Though computer system 10 is as specific specific descriptions, the present invention is defined as the computer system of demonstration 10, also can use on any electronic processing system.Further, as example, computing machine scale storage medium can comprise that much for instance, magnetic-based storage media is as disk (flexible plastic disc) or tape; Optical storage medium is as CD, optical magnetic tape or machine readable barcode; The solid-state electronic storage facilities, as random access memory (RAM), ROM (read-only memory) (ROM); Perhaps any other is used for the actual device or the medium of storage computation machine program.

Though this invention is set forth by top specific description, for those skilled in the art, can easily realize additional advantage and make amendment.Therefore its most widely aspect the present invention's examples shown of being not limited to specific details, representational equipment and illustrating here and describe, therefore under the situation of the spirit and scope of the general notion that does not deviate from attached claim and equivalency range and limited, can carry out various modifications.

Claims (6)

1, a kind of image of logarithmic code simultaneously method of carrying out denoising and sharpening is used for image is carried out denoising and sharpening synchronous processing, it is characterized in that, comprises following steps:

(a) from the luminance channel computational activity figure of digital image;

(b) use a table lookup correction activity diagram;

(c) the luminance channel of corrected activity diagram add-back digital image.

2, method according to claim 1 is characterized in that, comprises concrete steps:

(a) calculate luminance graph from digital image, it comprises the brightness value of digital pixel image;

(b) calculate activity diagram from luminance graph, the activity value of each pixel among its figure meets the activity around this pixel;

(c) if the activity magnitude of this pixel is the threshold value that is predetermined less than first, this activity then is judged as noise, revises with first predetermined adjustment earlier, is increased to the brightness value of this pixel then;

(d) if the activity magnitude of this pixel is the threshold value that is predetermined greater than first, less than second threshold value that is predetermined, this activity then is judged as based between noise and the edge, will be left intact to this pixel;

(e) if the activity magnitude of this pixel is the threshold value that is predetermined greater than second, less than the 3rd threshold value that is predetermined, it is weak edge that this activity then is judged as, and revises with second predetermined adjustment earlier, is increased to the brightness value of this pixel then;

(f) if the activity magnitude of this pixel is greater than the 3rd threshold value that is predetermined, it is strong edge that this activity then is judged as, and revises with the 3rd predetermined adjustment earlier, is increased to the brightness value of this pixel then.

3, method according to claim 2 is characterized in that, first predetermined adjustment is the denoising function.

4, method according to claim 2 is characterized in that, second predetermined adjustment is the sharpening function of appropriateness.

5, method according to claim 2 is characterized in that, the 3rd predetermined adjustment is the sharpening function of height.

6, method according to claim 2 is characterized in that, the adjustment of all activity diagram AM all realizes by lookup table LUT, and is as follows:

X is the activity magnitude of pixel, and x1, x2, x3 are respectively predetermined three threshold values.

Images