CN104657941A - Image edge self-adaptive enhancing method and device - Google Patents

Abstract

The invention discloses an image edge self-adaptive enhancing method and device and aims to solve the problems that the enhancement effect is common, a region which is not expected to be enhanced and noise are enhanced when an edge is enhanced, the calculated quantity is relatively large, the time consumption is relatively long and the instantaneity is poor in the prior art. The image edge self-adaptive enhancing method comprises the following steps: obtaining a detail index of each pixel point according to a neighborhood pixel point corresponding to each pixel point in an image; obtaining low, medium and high detail degrees of membership of each pixel point according to the detail index of each pixel point and a preset detail index threshold; denoising all pixel points with low detail degree of membership as 1, and performing edge enhancement on all pixel points with low detail degree of membership being less than 1 to obtain a final target image. Thus, details of each pixel point in the image can be pointedly enhanced, the noise is inhibited, no overshoot phenomenon is ensured; meanwhile, the image edge self-adaptive enhancing method is relatively small in calculated quantity in an implementing process, is reduced in calculating time, and has a better instantaneity.

Description

A kind of image border self-adapting enhancement method and device

Technical field

The present invention relates to image processing techniques, particularly relate to a kind of image border self-adapting enhancement method and device.

Background technology

The edge of image refers to the place that the local feature (as brightness, texture structure etc.) of image changes.Abundant information is contained in image border, has great significance to image processing field such as Iamge Segmentation, feature extraction, identifications.

Image border enhancing is the one of image enhancement processing.It is the technical method highlighted at edge (i.e. the boundary line of image tone sudden change or type of ground objects) place larger for the brightness value in image adjacent picture elements (or region) (or tone) difference.Image after image border strengthens more clearly can demonstrate the border of different type of ground objects or phenomenon or the trace of linear image, so that the identification of different type of ground objects and the delineation of distribution range thereof.

In prior art, have a lot of method to realize image border to strengthen, such as relevant mask technique and unsharp template etc., relevant mask technique is: copied by original image (image) and make a positive diaphragm and a negative diaphragm, and make two diaphragm precise overlay of different nature, when exposure is developed and printed, by distance very little for two the mutual changing of the relative positions of diaphragm, obtain the image that a corresponding image has misplace slightly " fringing " like this, the positive and negative counteracting of its most of image, and there is a bright line (or concealed wire) in its marginal portion, reach the display effect of outstanding image boundary line from background, image is made to reach enhancing.Unsharp template ratio juris is: first produce a passivation blurred picture by after original image low-pass filtering, original image and this blurred picture are subtracted each other the image obtaining reserved high-frequency composition, superpose with original image after being amplified by a parameter by high frequency imaging, this just produces the image that enhances edge again.

Traditional image border strengthens technology, strengthens effect general, needs user intervention image border technical process, often exists and strengthens edge simultaneously, and also enhance the region and noise of not wishing to strengthen, this is the common fault of many edge enhancing methods.

Summary of the invention

The embodiment of the present invention provides a kind of image border self-adapting enhancement method and device, strengthens effect generally, strengthens edge simultaneously, also enhance the region and noise of not wishing to strengthen in order to solve in prior art to exist.

The concrete technical scheme that the embodiment of the present invention provides is as follows:

First aspect, a kind of image border self-adapting enhancement method, comprising:

The neighborhood territory pixel point corresponding according to each pixel in image, obtains the details index of each pixel described, and wherein, described details index is the index weighing image border intensity;

According to the details index threshold that the details exponential sum of each pixel described is preset, obtain the low details degree of membership of each pixel described;

Carry out noise reduction process to the pixel that details degree of membership low in described image is 1, pixel details degree of membership low in described image being less than to 1 carries out edge enhancing process, obtains final target image.

Degree of membership is divided by the details index according to each pixel in image, and the pixel being 1 by low details degree of membership carries out filtering, other pixel is carried out edge and strengthen process, it is more natural that this degree of membership fuzzy division method makes image transition position process, and effectively can remove noise, strengthen the region needing to strengthen, strengthen the sharpness of image, and ensure that and there will not be overshoot phenomenon.

In conjunction with first aspect, in the implementation that the first is possible, when described image is coloured image, at the neighborhood territory pixel point corresponding according to each pixel in image, before obtaining the details index of each pixel described, also comprise:

Color space conversion process is carried out to described coloured image, described coloured image is converted to YC yuv space image;

And after obtaining final target image, also comprise:

Use following formula by the yuv space image after color space conversion process, revert to coloured image:

$\left\{\begin{array}{c}R=Y+1.403(V-128)\\ G=Y-0.344(U-128)-0.714(V-128)\\ B=Y+1.773(U-128)\end{array}\right.$

Wherein, Y represents the brightness of a pixel, U and V represents the colourity of a pixel, and R, G and B are the luminous intensity numerical value of red, green and blue in a pixel.

By this method, coloured image can be converted to and can carry out the yuv space image that image border self-adaptation strengthens process, and by yuv space image after treatment, coloured image can be reverted to.

In conjunction with the first possible implementation of first aspect or first aspect, in the implementation that the second is possible, the neighborhood territory pixel point corresponding according to any one pixel in described image, obtains the details index of any one pixel described, comprising:

Select (2n+1) * (2n+1) neighborhood in described image centered by any one pixel described, arithmetic mean calculating is carried out to the brightness value of all pixels in described neighborhood, obtain the neighborhood territory pixel average brightness that any one pixel described is corresponding, wherein, n be more than or equal to 1 positive integer;

By each pixel in described neighborhood square the adding up of difference of the brightness value neighborhood territory pixel average brightness corresponding with any one pixel described, obtain the details index of any one pixel described.

By this method, can the details index of each pixel in computed image accurately.

In conjunction with the first possible implementation of first aspect or first aspect, in the implementation that the third is possible, while the low details degree of membership obtaining each pixel described, also comprise:

According to the details index threshold that the details exponential sum of each pixel in described image is preset, obtain the middle details degree of membership of each pixel described and high details degree of membership.

By this method, obtain middle details degree of membership and high details degree of membership, just according to the difference of pixel details degree of membership, different enhancing process can be carried out.

In conjunction with the third possible implementation of first aspect, in the 4th kind of possible implementation, according to the details index threshold that the details exponential sum of any one pixel in described image is preset, obtain the low details degree of membership of any one pixel described, middle details degree of membership and high details degree of membership, comprising:

If the details index of any one pixel described is less than or equal to T ₁time, then the low details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the low details degree of membership of described arbitrary pixel and T ₂with the poor positive correlation of the details index of any one pixel described, with T ₂and T ₁poor negative correlation; If the details index of any one pixel described is more than or equal to T ₂time, then the low details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₁time, then the middle details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the middle details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₁poor positive correlation, with T ₂and T ₁poor negative correlation; If the details index of a described pixel is more than or equal to T ₂and be less than or equal to T ₃time, then the middle details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the middle details degree of membership of described arbitrary pixel and T ₄with the poor positive correlation of the details index of any one pixel described, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the middle details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₃time, then the high details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the high details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₃poor positive correlation, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the high details degree of membership of described arbitrary pixel is 1;

Wherein, T ₁, T ₂, T ₃and T ₄for the details index threshold preset, and meet 0 < T ₁< T ₂< T ₃< T ₄.

By this method, the low details degree of membership of each pixel in image, middle details degree of membership and high details degree of membership can be obtained accurately.

In conjunction with the first possible implementation of first aspect or first aspect, in the 5th kind of possible implementation, noise reduction process is carried out to the pixel that any one low details degree of membership in described image is 1, comprising:

Filter out in (2n+1) * (2n+1) neighborhood centered by the pixel that any one low details degree of membership described is 1, maximal value Maxf, the minimum M inf of neighborhood territory pixel brightness and intermediate value Medf, wherein, $\mathrm{Maxf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Max}}\left(f(x+i,y+j)\right),\mathrm{Minf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Min}}\left(f(x+i,y+j)\right),\mathrm{Medf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Median}}\left(f(x+i,y+j)\right);$

Medium filtering process is carried out to the pixel that any one low details degree of membership described is 1, if the brightness value that any one low details degree of membership described is the pixel of 1 is not equal to Maxf, and being not equal to Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 is constant; If the brightness value that any one low details degree of membership described is the pixel of 1 equals Maxf, or equals Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 equals Medf.

By this method, noise reduction process is carried out to the pixel that low details degree of membership is 1, effectively can remove noise.

In conjunction with the third possible implementation of first aspect, in the 6th kind of possible implementation, the pixel any one low details degree of membership in described image being less than to 1 carries out edge enhancing process, comprising:

Details degree of membership and high details degree of membership in the pixel of 1 is less than according to any one low details degree of membership described, and strengthen weight coefficient for the different edge that middle details degree of membership and high details degree of membership are preset, any one low details degree of membership described is less than details degree of membership and high details degree of membership in the pixel of 1 be weighted and calculate, obtains the edge enhancing weight coefficient that any one low details degree of membership described is less than the pixel of 1;

The brightness value and the Laplace sharpening template that any one low details degree of membership described are less than the pixel of 1 carry out convolution, obtain the skirt response value that any one low details degree of membership described is less than the pixel of 1;

By the brightness value any one low details degree of membership described being less than the pixel of 1 and the edge that any one low details degree of membership described is less than the pixel of 1 strengthen weight coefficient be multiplied with skirt response value after results added, the pixel being less than 1 to any one low details degree of membership described carries out edge and strengthens process, obtain pixel that any one low details degree of membership described is less than 1 carry out edge enhancing after pixel.

In this way, can to low details be subordinate to the pixel that is less than 1 according in details degree of membership and high details degree of membership carry out edge treated, effectively can carry out edge enhancing, strengthen the sharpness of image.

In conjunction with the first possible implementation of first aspect or first aspect, in the 7th kind of possible implementation, the different edge enhancing weight coefficient preset for middle details degree of membership and high details degree of membership should meet following condition: the edge enhancing weight coefficient preset for high details degree of membership is greater than 0, and is less than the edge enhancing weight coefficient preset for middle details degree of membership for the edge enhancing weight coefficient that high details degree of membership is preset.

In this way, according to the difference be subordinate to, different weight coefficients can be adopted to carry out edge targetedly to pixel and to strengthen process.

Second aspect, a kind of image border self-adaptation intensifier, comprising:

First computing unit, for the neighborhood territory pixel point corresponding according to each pixel in image, obtains the details index of each pixel described, and wherein, described details index is the index weighing image border intensity;

Second computing unit, for the details index threshold preset according to the details exponential sum of each pixel described, obtains the low details degree of membership of each pixel described;

Graphics processing unit, carries out noise reduction process for the pixel being 1 to details degree of membership low in described image, and pixel details degree of membership low in described image being less than to 1 carries out edge enhancing process, obtains final target image.

Like this, degree of membership is divided by the details index according to each pixel in image, and the pixel being 1 by low details degree of membership carries out filtering, other pixel is carried out edge and strengthen process, it is more natural that this degree of membership fuzzy division method makes image transition position process, and effectively can remove noise, strengthens the region needing to strengthen, strengthen the sharpness of image, and ensure that and there will not be overshoot phenomenon.

In conjunction with second aspect, in the implementation that the first is possible, when described image is coloured image, described image border self-adaptation intensifier, also comprises:

Converting unit, for at the first computing unit neighborhood territory pixel point corresponding according to each pixel in image, before obtaining the details index of each pixel described, color space conversion process is carried out to described coloured image, described coloured image is converted to YC yuv space image;

Recovery unit, after obtaining final target image at graphics processing unit, uses following formula by the yuv space image after color space conversion process, reverts to coloured image:

$\left\{\begin{array}{c}R=Y+1.403(V-128)\\ G=Y-0.344(U-128)-0.714(V-128)\\ B=Y+1.773(U-128)\end{array}\right.$

Wherein, Y represents the brightness of a pixel, U and V represents the colourity of a pixel, and R, G and B are the luminous intensity numerical value of red, green and blue in a pixel.

Like this, coloured image can be converted to and can carry out the yuv space image that image border self-adaptation strengthens process by conversion unit, and recovery unit by yuv space image after treatment, can revert to coloured image.。

In conjunction with the first possible implementation of second aspect or second aspect, in the implementation that the second is possible, the neighborhood territory pixel point that described first computing unit is corresponding according to any one pixel in described image, obtains the details index of any one pixel described, comprising:

Described first computing unit selects (2n+1) * (2n+1) neighborhood in described image centered by any one pixel described, arithmetic mean calculating is carried out to the brightness value of all pixels in described neighborhood, obtain the neighborhood territory pixel average brightness that any one pixel described is corresponding, wherein, n be more than or equal to 1 positive integer;

Described first computing unit by each pixel in described neighborhood square the adding up of difference of the brightness value neighborhood territory pixel average brightness corresponding with any one pixel described, obtain the details index of any one pixel described.

Like this, the first computing unit can the details index of each pixel in computed image accurately.

In conjunction with the first possible implementation of second aspect or second aspect, in the implementation that the third is possible, obtain the low details degree of membership of each pixel described at described second computing unit while, also comprise:

The details index threshold that described second computing unit is preset according to the details exponential sum of each pixel in described image, obtains the middle details degree of membership of each pixel described and high details degree of membership.

Like this, the second computing unit obtains middle details degree of membership and high details degree of membership, just according to the difference of pixel details degree of membership, can carry out different enhancing process.

In conjunction with the third possible implementation of second aspect, in the 4th kind of possible implementation, the details index threshold that described second computing unit is preset according to the details exponential sum of any one pixel in described image, obtain the low details degree of membership of any one pixel described, middle details degree of membership and high details degree of membership, comprising:

If the details index of any one pixel described is less than or equal to T ₁time, then the low details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the low details degree of membership of described arbitrary pixel and T ₂with the poor positive correlation of the details index of any one pixel described, with T ₂and T ₁poor negative correlation; If the details index of any one pixel described is more than or equal to T ₂time, then the low details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₁time, then the middle details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the middle details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₁poor positive correlation, with T ₂and T ₁poor negative correlation; If the details index of a described pixel is more than or equal to T ₂and be less than or equal to T ₃time, then the middle details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the middle details degree of membership of described arbitrary pixel and T ₄with the poor positive correlation of the details index of any one pixel described, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the middle details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₃time, then the high details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the high details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₃poor positive correlation, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the high details degree of membership of described arbitrary pixel is 1;

Wherein, T ₁, T ₂, T ₃and T ₄for the details index threshold preset, and meet 0 < T ₁< T ₂< T ₃< T ₄.

Like this, the second computing unit can obtain the low details degree of membership of each pixel in image, middle details degree of membership and high details degree of membership accurately.

In conjunction with the first possible implementation of second aspect or second aspect, in the 5th kind of possible implementation, described graphics processing unit carries out noise reduction process to the pixel that any one low details degree of membership in described image is 1, comprising:

Described graphics processing unit filters out in (2n+1) * (2n+1) neighborhood centered by the pixel that any one low details degree of membership described is 1, maximal value Maxf, the minimum M inf of neighborhood territory pixel brightness and intermediate value Medf, wherein, $\mathrm{Maxf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Max}}\left(f(x+i,y+j)\right),\mathrm{Minf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Min}}\left(f(x+i,y+j)\right),$ $\mathrm{Medf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Median}}\left(f(x+i,y+j)\right);$

Described graphics processing unit carries out medium filtering process to the pixel that any one low details degree of membership described is 1, if the brightness value that any one low details degree of membership described is the pixel of 1 is not equal to Maxf, and being not equal to Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 is constant; If the brightness value that any one low details degree of membership described is the pixel of 1 equals Maxf, or equals Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 equals Medf.

Like this, graphics processing unit carries out noise reduction process to the pixel that low details degree of membership is 1, effectively can remove noise.

In conjunction with the third possible implementation of second aspect, in the 6th kind of possible implementation, described graphics processing unit carries out edge enhancing process to the pixel that any one low details degree of membership in described image is less than 1, comprising:

Described graphics processing unit is less than the middle details degree of membership of the pixel of 1 and high details degree of membership according to any one low details degree of membership described, and strengthen weight coefficient for the different edge that middle details degree of membership and high details degree of membership are preset, any one low details degree of membership described is less than details degree of membership and high details degree of membership in the pixel of 1 be weighted and calculate, obtains the edge enhancing weight coefficient that any one low details degree of membership described is less than the pixel of 1;

Any one low details degree of membership described is less than the brightness value of the pixel of 1 by described graphics processing unit and Laplace sharpening template carries out convolution, obtains the skirt response value that any one low details degree of membership described is less than the pixel of 1;

Described graphics processing unit by the brightness value any one low details degree of membership described being less than the pixel of 1 and the edge that any one low details degree of membership described is less than the pixel of 1 strengthen weight coefficient be multiplied with skirt response value after results added, the pixel being less than 1 to any one low details degree of membership described carries out edge and strengthens process, obtain pixel that any one low details degree of membership described is less than 1 carry out edge enhancing after pixel.

Like this, graphics processing unit can to low details be subordinate to the pixel that is less than 1 according in details degree of membership and high details degree of membership carry out edge treated, effectively can carry out edge enhancing, strengthen the sharpness of image.

In conjunction with the first possible implementation of second aspect or second aspect, in the 7th kind of possible implementation, the different edge enhancing weight coefficient preset for middle details degree of membership and high details degree of membership should meet following condition: the edge enhancing weight coefficient preset for high details degree of membership is greater than 0, and is less than the edge enhancing weight coefficient preset for middle details degree of membership for the edge enhancing weight coefficient that high details degree of membership is preset.

Like this, graphics processing unit according to the difference be subordinate to, can adopt different weight coefficients to carry out edge targetedly to pixel and strengthens process.

Adopt technical solution of the present invention, user intervention edge is not needed to strengthen process, degree of membership is divided by the details index according to each pixel in image, and the pixel being 1 by low details degree of membership carries out filtering, other pixel is carried out edge and strengthen process, it is more natural that this degree of membership fuzzy division method makes image transition position process, can strengthen targetedly according to the details of pixel each in image, inhibit noise simultaneously, guarantee there will not be overshoot phenomenon, and calculated amount is less in technical solution of the present invention implementation process, save computing time, real-time is good, can effectively avoid the enhancing effect that exists in prior art general, strengthen edge simultaneously, also enhance the problem of not wishing region and the noise strengthened.

Accompanying drawing explanation

The particular flow sheet of a kind of image border self-adapting enhancement method that Fig. 1 provides for the embodiment of the present invention;

The schematic diagram choosing (2n+1) * (2n+1) neighborhood that Fig. 2 provides for the embodiment of the present invention;

The details degree of membership curve synoptic diagram that Fig. 3 provides for the embodiment of the present invention;

The preferred Laplace sharpening template schematic diagram that Fig. 4 provides for the embodiment of the present invention;

The schematic diagram of a kind of image border self-adaptation intensifier that Fig. 5 provides for the embodiment of the present invention.

Embodiment

Adopt technical solution of the present invention, user intervention edge is not needed to strengthen process, can strengthen targetedly according to the details of pixel each in image, inhibit noise, and in technical solution of the present invention implementation process, calculated amount is less, saves computing time simultaneously, real-time is good, the enhancing effect that exists in prior art effectively can be avoided general, strengthen edge simultaneously, also enhance the problem of not wishing region and the noise strengthened.

Embodiments provide a kind of image border self-adapting enhancement method, below in conjunction with accompanying drawing, the preferred embodiment of the present invention is described in detail.

Consult shown in Fig. 1, the concrete treatment scheme of image border self-adapting enhancement method comprises:

Step 101: the neighborhood territory pixel point corresponding according to each pixel in image, obtains the details index of each pixel in this image, wherein, details index is the index weighing image border intensity.

The embodiment of the present invention can realize carrying out edge enhancing to gray level image and coloured image, the image border self-adapting enhancement method provided due to the embodiment of the present invention is only at brightness (Luma, Y) plane is carried out, if desired the image strengthened is coloured image, then before carry out step 101, first need this coloured image to carry out color space conversion process, be converted to YC (LumaChroma, YUV) spatial image.

Particularly, if original image is RGB (RedGreenBlue, RGB) spatial image, then needs rgb format image to be converted to yuv space image, see shown in formula one.

$\left\{\begin{array}{c}Y=0.299R+0.587G+0.114B\\ U=0.564(B-Y)+128\\ V=0.713(R-Y)+128\end{array}\right.$ Formula one

Wherein, Y represents the brightness of a pixel, U and V represents the colourity of a pixel, and describe color and the saturation degree of a pixel, R, G and B are the luminous intensity numerical value of red, green and blue in a pixel.

In practice process, formula rgb format image being converted to yuv space image has a lot, such as, can also adopt $\left\{\begin{array}{c}Y=0.257R+0.504G+0.098B+16\\ U=-0.148R-0.291G+0.439B+128\\ V=0.439R-0.368G-0.071B+128\end{array}\right.$ Change, preferably, the embodiment of the present invention selects formula one to carry out color space conversion process.

The neighborhood territory pixel point corresponding according to any one pixel in image, obtains the details index of any one pixel in this image, specifically comprises:

First, select (2n+1) * (2n+1) neighborhood in this image centered by this any one pixel, wherein, n be more than or equal to 1 positive integer.

Such as, consult shown in Fig. 2, in 201, C is this any one pixel, * be the pixel in (2n+1) * (2n+1) neighborhood in this image centered by this any one pixel, if choose in the process of (2n+1) * (2n+1) neighborhood centered by this any one pixel in the images, occur that this any one pixel is the pixel in image border, neighborhood territory pixel point is chosen at the relative edge of this any one image border, pixel place, such as, consult shown in 202, wherein, n is 1.

Then, arithmetic mean calculating is carried out to the brightness value of all pixels in this neighborhood, obtain the neighborhood territory pixel average brightness that this any one pixel is corresponding.

Using formula two, carries out arithmetic mean calculating to the brightness value of all pixels in this neighborhood:

$\stackrel{\&OverBar;}{f}(x,y)=\frac{1}{{(2n+1)}^2}\underset{i=x-n}{\overset{x+n}{\mathrm{\&Sigma;}}}\underset{j=y-n}{\overset{y+n}{\mathrm{\&Sigma;}}}f(i,j)$ Formula two

Wherein, represent the neighborhood territory pixel average brightness that this any one pixel is corresponding, x represents the horizontal ordinate of this any one pixel, and y represents the ordinate of this any one pixel, n represents the positive integer being more than or equal to 1, f (i, j) is the brightness value of the pixel of (i, j) for coordinate;

Finally, by each pixel in this neighborhood square the adding up of difference of neighborhood territory pixel average brightness corresponding to any one pixel of brightness value and this, see formula three, obtain the details index of this any one pixel:

$D(x,y)=\frac{1}{{(2n+1)}^2}\underset{i=x-n}{\overset{x+n}{\mathrm{\&Sigma;}}}\underset{j=y-n}{\overset{y+n}{\mathrm{\&Sigma;}}}{(f(i,j)-\stackrel{\&OverBar;}{f}(x,y))}^2$ Formula three

Wherein, D (x, y) represents the details index of this any one pixel.

In embodiments of the present invention, in order to improve operation efficiency, avoid the division arithmetic in operational process, preferably as far as possible, carry out arithmetic mean calculating by the brightness value of formula four to all pixels in this neighborhood, calculated the details index of any one pixel by formula five:

$\stackrel{\&OverBar;}{f}(x,y)=\underset{i=x-n}{\overset{x+n}{\mathrm{\&Sigma;}}}\underset{j=y-n}{\overset{y+n}{\mathrm{\&Sigma;}}}f(i,j)$ Formula four

$D(x,y)=\underset{i=x-n}{\overset{x+n}{\mathrm{\&Sigma;}}}\underset{j=y-n}{\overset{y+n}{\mathrm{\&Sigma;}}}{({(2n+1)}^2*(f(i,j)-\stackrel{\&OverBar;}{f}(x,y)))}^2$ Formula five

Step 102: the details index threshold preset according to the details exponential sum of each pixel in this image, obtains the low details degree of membership of each pixel of this image;

According to the details index threshold that the details exponential sum of each pixel in this image is preset, while obtaining the low details degree of membership of each pixel of this image, the middle details degree of membership of each pixel of this image and high details degree of membership can also be obtained.

Particularly, according to the details index threshold that the details exponential sum of any one pixel in this image is preset, obtain the low details degree of membership of this any one pixel, middle details degree of membership and high details degree of membership, comprising:

If the details index of this any one pixel is less than or equal to T ₁time, then the low details degree of membership of this arbitrary pixel is 1; If the details index of this any one pixel is greater than T ₁and be less than T ₂time, then the low details degree of membership of this arbitrary pixel and T ₂with the poor positive correlation of the details index of this any one pixel, with T ₂and T ₁poor negative correlation; If the details index of this any one pixel is more than or equal to T ₂time, then the low details degree of membership of this arbitrary pixel is 0;

If the details index of this any one pixel is less than or equal to T ₁time, then the middle details degree of membership of this arbitrary pixel is 0; If the details index of this any one pixel is greater than T ₁and be less than T ₂time, then the middle details degree of membership of this arbitrary pixel and the details exponential sum T of this any one pixel ₁poor positive correlation, with T ₂and T ₁poor negative correlation; If the details index of this pixel is more than or equal to T ₂and be less than or equal to T ₃time, then the middle details degree of membership of this arbitrary pixel is 1; If the details index of this any one pixel is greater than T ₃and be less than T ₄time, then the middle details degree of membership of this arbitrary pixel and T ₄with the poor positive correlation of the details index of this any one pixel, with T ₄and T ₃poor negative correlation; If the details index of this any one pixel is more than or equal to T ₄time, then the middle details degree of membership of this arbitrary pixel is 0;

If the details index of this any one pixel is less than or equal to T ₃time, then the high details degree of membership of this arbitrary pixel is 0; If the details index of this any one pixel is greater than T ₃and be less than T ₄time, then the high details degree of membership of this arbitrary pixel and the details exponential sum T of this any one pixel ₃poor positive correlation, with T ₄and T ₃poor negative correlation; If the details index of this any one pixel is more than or equal to T ₄time, then the high details degree of membership of this arbitrary pixel is 1;

Wherein, T ₁, T ₂, T ₃and T ₄for the details index threshold preset, and meet 0 < T ₁< T ₂< T ₃< T ₄.

Preferably, in embodiments of the present invention, the low details degree of membership of each pixel of this image, middle details degree of membership and high details degree of membership can be calculated by formula six, formula seven and formula eight:

${\mathrm{\&mu;}}_1\left(D(x,y)\right)=\left\{\begin{array}{cc}1& D(x,y)\&le;T_1\\ \frac{T_2-D(x,y)}{T_2-T_1}& T_1<D(x,y)<T_2\\ 0& D(x,y)\&GreaterEqual;T_2\end{array}\right.$ Formula six

${\mathrm{\&mu;}}_2\left(D(x,y)\right)=\left\{\begin{array}{cc}0& D(x,y)\&le;T_1\\ \frac{D(x,y)-T_1}{T_2-T_1}& T_1<D(x,y)<T_2\\ 1& T_2\&le;D(x,y)\&le;T_3\\ \frac{T_4-D(x,y)}{T_4-T_3}& T_3<D(x,y)<T_4\\ 0& D(x,y)\&GreaterEqual;T_4\end{array}\right.$ Formula seven

${\mathrm{\&mu;}}_3\left(D(x,y)\right)=\left\{\begin{array}{cc}0& D(x,y)\&le;T_3\\ \frac{D(x,y)-T_3}{T_4-T_3}& T_3<D(x,y)<T_4\\ 1& D(x,y)\&GreaterEqual;T_4\end{array}\right.$ Formula eight

Wherein, μ ₁(D (x, y)), μ ₂(D (x, y)) and μ ₃(D (x, y)) is respectively the low details degree of membership of this any one pixel, middle details degree of membership and high details degree of membership, and its details membership function curve is consulted shown in Fig. 3.

Each pixel in image has corresponding low details degree of membership, middle details degree of membership and high details degree of membership, and the low details degree of membership that in image, each pixel is corresponding, middle details degree of membership and high details degree of membership sum are 1.

Step 103: carry out noise reduction process to the pixel that details degree of membership low in this image is 1, pixel details degree of membership low in this image being less than to 1 carries out edge enhancing process, obtains final target image.

Particularly, noise reduction process is carried out to the pixel that any one low details degree of membership in all pixels of this image is 1, comprising:

Filter out any one low details degree of membership by this be 1 pixel centered by (2n+1) * (2n+1) neighborhood in, maximal value Maxf, the minimum M inf of neighborhood territory pixel brightness and intermediate value Medf, wherein, $\mathrm{Maxf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Max}}\left(f(x+i,y+j)\right),\mathrm{Minf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Min}}\left(f(x+i,y+j)\right),\mathrm{Medf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Median}}\left(f(x+i,y+j)\right),$ Intermediate value Medf is that any one low details degree of membership is (2n+1) * (2n+1) neighborhood (2n+1) that comprise centered by the pixel of 1 by this ²individual pixel sorts according to brightness value size, and selected and sorted is the brightness value of middle pixel.

To this any one low details degree of membership be 1 pixel carry out medium filtering process, if the brightness value that this any one low details degree of membership is the pixel of 1 is not equal to Maxf, and being not equal to Minf, then this any one low details degree of membership is that the brightness value of the pixel of 1 is constant; If the brightness value that this any one low details degree of membership is the pixel of 1 equals Maxf, or equals Minf, then this any one low details degree of membership is that the brightness value of the pixel of 1 equals Medf.

Preferably, can using formula nine to this any one low details degree of membership be 1 pixel carry out medium filtering process:

formula nine

Particularly, the pixel any one low details degree of membership in all pixels of this image being less than to 1 carries out edge enhancing process, comprising:

First, according to this, any one low details degree of membership is less than details degree of membership and high details degree of membership in the pixel of 1, and strengthen weight coefficient for the different edge that middle details degree of membership and high details degree of membership are preset, by this, any one low details degree of membership is less than details degree of membership and high details degree of membership in the pixel of 1 and is weighted and calculates, and obtains the edge that this any one low details degree of membership is less than the pixel of 1 and strengthens weight coefficient.

The different edge enhancing weight coefficient preset for middle details degree of membership and high details degree of membership is k ₁and k ₂, wherein, 0 < k ₂< k ₁, preferably, can using formula ten calculate this any one low details degree of membership be less than the pixel of 1 edge strengthen weight coefficient:

K (x, y)=μ ₂(D (x, y)) * k ₁+ μ ₃(D (x, y)) * k ₂formula ten

The edge that k (x, y) is less than the pixel of 1 for this any one low details degree of membership strengthens weight coefficient.

Then, by this, any one low details degree of membership is less than the brightness value of the pixel of 1 and Laplace sharpening template carries out convolution, obtains the skirt response value that this any one low details degree of membership is less than the pixel of 1.

In actual applications, calculate the skirt response value that this any one low details degree of membership is less than the pixel of 1, usually adopt unsharp template and Laplace sharpening method, the present invention selects Laplace sharpening method better to realize image border self-adaptation enhancing effect.

Preferably, the skirt response value that this any one low details degree of membership is less than the pixel of 1 can be obtained by using formula 11:

$f_{\mathrm{Laplace}}(x,y)=f(x,y)\&CircleTimes;L_p$ Formula 11

Wherein, f _laplace(x, y) is less than the skirt response value of the pixel of 1, L for this any one low details degree of membership _pfor Laplace sharpening template.

L _pfor the Filtering Template of (2m+1) * (2m+1), wherein, m is positive integer, consults shown in Fig. 4, in the embodiment of the present invention, and the L of preferred 3*3 _p.

Finally, the edge that any one the low details degree of membership of brightness value and this being less than the pixel of 1 by any one low details degree of membership by this is less than the pixel of 1 strengthen weight coefficient be multiplied with skirt response value after results added, the pixel that any one low details degree of membership is less than 1 to this carries out edge and strengthens process, obtain pixel that this any one low details degree of membership is less than 1 carry out edge enhancing after pixel.

According to formula ten and formula 11, embodiment of the present invention preferred formula 12, the pixel that any one low details degree of membership is less than 1 to this carries out edge and strengthens process:

G (x, y)=f (x, y)+k (x, y) * f _laplace(x, y) formula 12

Wherein, the pixel that g (x, y) is less than 1 for this any one low details degree of membership carries out the pixel after edge enhancing.

Noise reduction process is carried out to the pixel that low details degrees of membership all in all pixels of this image are 1, the pixel low details degrees of membership all in all pixels of this image being less than to 1 carries out edge enhancing process, obtain final target image F (x, y), see formula 13.

formula 13

Obtain final target image F (x, y) for image border self-adaptation strengthen after image, if original image is coloured image, after obtaining F (x, y), also needing this image from yuv space Graphic Exchanging is coloured image, sees formula 14:

$\left\{\begin{array}{c}R=Y+1.403(V-128)\\ G=Y-0.344(U-128)-0.714(V-128)\\ B=Y+1.773(U-128)\end{array}\right.$ Formula 14

Based on above-described embodiment, consult shown in Fig. 5, the embodiment of the present invention additionally provides a kind of image border self-adaptation intensifier, comprising: the first computing unit 501, second computing unit 502 and graphics processing unit 503, wherein

First computing unit 501, for the neighborhood territory pixel point corresponding according to each pixel in image, obtains the details index of each pixel in this image, and wherein, details index is the index weighing image border intensity;

Second computing unit 502, for the details index threshold preset according to the details exponential sum of each pixel in this image, obtains the low details degree of membership of each pixel of this image;

Graphics processing unit 503, carries out noise reduction process for the pixel being 1 to details degree of membership low in this image, and pixel details degree of membership low in this image being less than to 1 carries out edge enhancing process, obtains final target image.

When this image is coloured image, image border self-adaptation intensifier, also comprises:

Converting unit 500, for at the first computing unit neighborhood territory pixel point corresponding according to each pixel in image, before obtaining the details index of each pixel in this image, color space conversion process is carried out to this coloured image, this coloured image is converted to YC yuv space image;

Recovery unit 504, after obtaining final target image at graphics processing unit, uses following formula by the yuv space image after color space conversion process, reverts to coloured image:

$\left\{\begin{array}{c}R=Y+1.403(V-128)\\ G=Y-0.344(U-128)-0.714(V-128)\\ B=Y+1.773(U-128)\end{array}\right.$

Wherein, Y represents the brightness of a pixel, U and V represents the colourity of a pixel, and R, G and B are the luminous intensity numerical value of red, green and blue in a pixel.

The neighborhood territory pixel point that first computing unit 501 is corresponding according to any one pixel in this image, obtains the details index of any one pixel, comprising:

First computing unit 501 selects (2n+1) * (2n+1) neighborhood in this image centered by any one pixel, arithmetic mean calculating is carried out to the brightness value of all pixels in this neighborhood, obtains the neighborhood territory pixel average brightness that this any one pixel is corresponding;

First computing unit 501 by each pixel in this neighborhood square the adding up of difference of neighborhood territory pixel average brightness corresponding to any one pixel of brightness value and this, obtain the details index of this any one pixel.

Obtain the low details degree of membership of each pixel in this image at the second computing unit 502 while, also comprise:

The details index threshold that second computing unit 502 is preset according to the details exponential sum of each pixel in this image, obtains the middle details degree of membership of each pixel in this image and high details degree of membership.

The details index threshold that second computing unit 502 is preset according to the details exponential sum of any one pixel in this image, obtains the low details degree of membership of this any one pixel, middle details degree of membership and high details degree of membership, comprising:

If the details index of this any one pixel is less than or equal to T ₁time, then the low details degree of membership of this arbitrary pixel is 1; If the details index of this any one pixel is greater than T ₁and be less than T ₂time, then the low details degree of membership of this arbitrary pixel and T ₂with the poor positive correlation of the details index of this any one pixel, with T ₂and T ₁poor negative correlation; If the details index of this any one pixel is more than or equal to T ₂time, then the low details degree of membership of this arbitrary pixel is 0;

If the details index of this any one pixel is less than or equal to T ₁time, then the middle details degree of membership of this arbitrary pixel is 0; If the details index of this any one pixel is greater than T ₁and be less than T ₂time, then the middle details degree of membership of this arbitrary pixel and the details exponential sum T of this any one pixel ₁poor positive correlation, with T ₂and T ₁poor negative correlation; If the details index of this pixel is more than or equal to T ₂and be less than or equal to T ₃time, then the middle details degree of membership of this arbitrary pixel is 1; If the details index of this any one pixel is greater than T ₃and be less than T ₄time, then the middle details degree of membership of this arbitrary pixel and T ₄with the poor positive correlation of the details index of this any one pixel, with T ₄and T ₃poor negative correlation; If the details index of this any one pixel is more than or equal to T ₄time, then the middle details degree of membership of this arbitrary pixel is 0;

If the details index of this any one pixel is less than or equal to T ₃time, then the high details degree of membership of this arbitrary pixel is 0; If the details index of this any one pixel is greater than T ₃and be less than T ₄time, then the high details degree of membership of this arbitrary pixel and the details exponential sum T of this any one pixel ₃poor positive correlation, with T ₄and T ₃poor negative correlation; If the details index of this any one pixel is more than or equal to T ₄time, then the high details degree of membership of this arbitrary pixel is 1;

Wherein, T ₁, T ₂, T ₃and T ₄for the details index threshold preset, and meet 0 < T ₁< T ₂< T ₃< T ₄.

Graphics processing unit 503 carries out noise reduction process to the pixel that any one low details degree of membership in all pixels of this image is 1, comprising:

Graphics processing unit 503 filter out any one low details degree of membership by this be 1 pixel centered by (2n+1) * (2n+1) neighborhood in, maximal value Maxf, the minimum M inf of neighborhood territory pixel brightness and intermediate value Medf, wherein, $\mathrm{Maxf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Max}}\left(f(x+i,y+j)\right),\mathrm{Minf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Min}}\left(f(x+i,y+j)\right),$ $\mathrm{Medf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Median}}\left(f(x+i,y+j)\right);$

Graphics processing unit 503 to this any one low details degree of membership be 1 pixel carry out medium filtering process, if the brightness value that this any one low details degree of membership is the pixel of 1 is not equal to Maxf, and being not equal to Minf, then this any one low details degree of membership is that the brightness value of the pixel of 1 is constant; If the brightness value that this any one low details degree of membership is the pixel of 1 equals Maxf, or equals Minf, then this any one low details degree of membership is that the brightness value of the pixel of 1 equals Medf.

Graphics processing unit 503 carries out edge enhancing process to the pixel that any one low details degree of membership in all pixels of this image is less than 1, comprising:

According to this, any one low details degree of membership is less than details degree of membership and high details degree of membership in the pixel of 1 to graphics processing unit 503, and strengthen weight coefficient for the different edge that middle details degree of membership and high details degree of membership are preset, by this, any one low details degree of membership is less than details degree of membership and high details degree of membership in the pixel of 1 and is weighted and calculates, and obtains the edge that this any one low details degree of membership is less than the pixel of 1 and strengthens weight coefficient;

By this, any one low details degree of membership is less than the brightness value of the pixel of 1 and Laplace sharpening template carries out convolution to graphics processing unit 503, obtains the skirt response value that this any one low details degree of membership is less than the pixel of 1;

Graphics processing unit 503 by the edge that any one low details degree of membership of brightness value and this that any one low details degree of membership by this is less than the pixel of 1 is less than the pixel of 1 strengthen weight coefficient be multiplied with skirt response value after results added, the pixel that any one low details degree of membership is less than 1 to this carries out edge and strengthens process, obtain pixel that this any one low details degree of membership is less than 1 carry out edge enhancing after pixel.

The different edge enhancing weight coefficient preset for middle details degree of membership and high details degree of membership should meet following condition: the edge enhancing weight coefficient preset for high details degree of membership is greater than 0, and is less than the different edge enhancing weight coefficient preset for middle details degree of membership for the edge enhancing weight coefficient that high details degree of membership is preset.

In sum, a kind of image border self-adapting enhancement method provided by the embodiment of the present invention and device, user intervention edge is not needed to strengthen process, degree of membership is divided by the details index according to each pixel in image, and the pixel being 1 by low details degree of membership carries out filtering, other pixel is carried out edge and strengthen process, it is more natural that this degree of membership fuzzy division method makes image transition position process, can strengthen targetedly according to the details of pixel each in image, inhibit noise simultaneously, guarantee there will not be overshoot phenomenon, and calculated amount is less in technical solution of the present invention implementation process, save computing time, real-time is good, can effectively avoid the enhancing effect that exists in prior art general, strengthen edge simultaneously, also enhance the problem of not wishing region and the noise strengthened.

Although describe the preferred embodiments of the present invention, those skilled in the art once obtain the basic creative concept of cicada, then can make other change and amendment to these embodiments.So claims are intended to be interpreted as comprising preferred embodiment and falling into all changes and the amendment of the scope of the invention.

Obviously, those skilled in the art can carry out various change and modification to the embodiment of the present invention and not depart from the spirit and scope of the embodiment of the present invention.Like this, if these amendments of the embodiment of the present invention and modification belong within the scope of the claims in the present invention and equivalent technologies thereof, then the present invention is also intended to comprise these change and modification.

Claims (16)

1. an image border self-adapting enhancement method, is characterized in that, comprising:

The neighborhood territory pixel point corresponding according to each pixel in image, obtains the details index of each pixel described, and wherein, described details index is the index weighing image border intensity;

According to the details index threshold that the details exponential sum of each pixel described is preset, obtain the low details degree of membership of each pixel described;

Carry out noise reduction process to the pixel that details degree of membership low in described image is 1, pixel details degree of membership low in described image being less than to 1 carries out edge enhancing process, obtains final target image.

2. the method for claim 1, is characterized in that, when described image is coloured image, at the neighborhood territory pixel point corresponding according to each pixel in image, before obtaining the details index of each pixel described, also comprises:

Color space conversion process is carried out to described coloured image, described coloured image is converted to YC yuv space image;

And after obtaining final target image, also comprise:

Use following formula by the yuv space image after color space conversion process, revert to coloured image:

$\left\{\begin{array}{c}R=Y+1.403(V-128)\\ G=Y-0.344(U-128)-0.714(V-128)\\ B=Y+1.773(U-128)\end{array}\right.$

Wherein, Y represents the brightness of a pixel, U and V represents the colourity of a pixel, and R, G and B are the luminous intensity numerical value of red, green and blue in a pixel.

3. method as claimed in claim 1 or 2, is characterized in that, the neighborhood territory pixel point corresponding according to any one pixel in described image, obtains the details index of any one pixel described, comprising:

Select (2n+1) * (2n+1) neighborhood in described image centered by any one pixel described, arithmetic mean calculating is carried out to the brightness value of all pixels in described neighborhood, obtain the neighborhood territory pixel average brightness that any one pixel described is corresponding, wherein, n be more than or equal to 1 positive integer;

By each pixel in described neighborhood square the adding up of difference of the brightness value neighborhood territory pixel average brightness corresponding with any one pixel described, obtain the details index of any one pixel described.

4. method as claimed in claim 1 or 2, is characterized in that, while the low details degree of membership obtaining each pixel described, also comprises:

According to the details index threshold that the details exponential sum of each pixel in described image is preset, obtain the middle details degree of membership of each pixel described and high details degree of membership.

5. method as claimed in claim 4, it is characterized in that, according to the details index threshold that the details exponential sum of any one pixel in described image is preset, obtain the low details degree of membership of any one pixel described, middle details degree of membership and high details degree of membership, comprising:

If the details index of any one pixel described is less than or equal to T ₁time, then the low details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the low details degree of membership of described arbitrary pixel and T ₂with the poor positive correlation of the details index of any one pixel described, with T ₂and T ₁poor negative correlation; If the details index of any one pixel described is more than or equal to T ₂time, then the low details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₁time, then the middle details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the middle details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₁poor positive correlation, with T ₂and T ₁poor negative correlation; If the details index of a described pixel is more than or equal to T ₂and be less than or equal to T ₃time, then the middle details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the middle details degree of membership of described arbitrary pixel and T ₄with the poor positive correlation of the details index of any one pixel described, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the middle details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₃time, then the high details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the high details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₃poor positive correlation, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the high details degree of membership of described arbitrary pixel is 1;

Wherein, T ₁, T ₂, T ₃and T ₄for the details index threshold preset, and meet 0 < T ₁< T ₂< T ₃< T ₄.

6. method as claimed in claim 1 or 2, is characterized in that, carry out noise reduction process, comprising the pixel that any one low details degree of membership in described image is 1:

Filter out in (2n+1) * (2n+1) neighborhood centered by the pixel that any one low details degree of membership described is 1, maximal value Maxf, the minimum M inf of neighborhood territory pixel brightness and intermediate value Medf, wherein, $\mathrm{Maxf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Max}}\left(f(x+i,y+j)\right),\mathrm{Minf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Min}}\left(f(x+i,y+j)\right),\mathrm{Medf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Median}}\left(f(x+i,y+j)\right);$

Medium filtering process is carried out to the pixel that any one low details degree of membership described is 1, if the brightness value that any one low details degree of membership described is the pixel of 1 is not equal to Maxf, and being not equal to Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 is constant; If the brightness value that any one low details degree of membership described is the pixel of 1 equals Maxf, or equals Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 equals Medf.

7. method as claimed in claim 4, is characterized in that, the pixel any one low details degree of membership in described image being less than to 1 carries out edge enhancing process, comprising:

Details degree of membership and high details degree of membership in the pixel of 1 is less than according to any one low details degree of membership described, and strengthen weight coefficient for the different edge that middle details degree of membership and high details degree of membership are preset, any one low details degree of membership described is less than details degree of membership and high details degree of membership in the pixel of 1 be weighted and calculate, obtains the edge enhancing weight coefficient that any one low details degree of membership described is less than the pixel of 1;

The brightness value and the Laplace sharpening template that any one low details degree of membership described are less than the pixel of 1 carry out convolution, obtain the skirt response value that any one low details degree of membership described is less than the pixel of 1;

By the brightness value any one low details degree of membership described being less than the pixel of 1 and the edge that any one low details degree of membership described is less than the pixel of 1 strengthen weight coefficient be multiplied with skirt response value after results added, the pixel being less than 1 to any one low details degree of membership described carries out edge and strengthens process, obtain pixel that any one low details degree of membership described is less than 1 carry out edge enhancing after pixel.

8. method as claimed in claim 7, it is characterized in that, the different edge enhancing weight coefficient preset for middle details degree of membership and high details degree of membership should meet following condition: the edge enhancing weight coefficient preset for high details degree of membership is greater than 0, and is less than the edge enhancing weight coefficient preset for middle details degree of membership for the edge enhancing weight coefficient that high details degree of membership is preset.

9. an image border self-adaptation intensifier, is characterized in that, comprising:

First computing unit, for the neighborhood territory pixel point corresponding according to each pixel in image, obtains the details index of each pixel described, and wherein, described details index is the index weighing image border intensity;

Second computing unit, for the details index threshold preset according to the details exponential sum of each pixel described, obtains the low details degree of membership of each pixel described;

Graphics processing unit, carries out noise reduction process for the pixel being 1 to details degree of membership low in described image, and pixel details degree of membership low in described image being less than to 1 carries out edge enhancing process, obtains final target image.

10. device as claimed in claim 9, it is characterized in that, when described image is coloured image, described image border self-adaptation intensifier, also comprises:

Converting unit, for at the first computing unit neighborhood territory pixel point corresponding according to each pixel in image, before obtaining the details index of each pixel described, color space conversion process is carried out to described coloured image, described coloured image is converted to YC yuv space image;

Recovery unit, after obtaining final target image at graphics processing unit, uses following formula by the yuv space image after color space conversion process, reverts to coloured image:

$\left\{\begin{array}{c}R=Y+1.403(V-128)\\ G=Y-0.344(U-128)-0.714(V-128)\\ B=Y+1.773(U-128)\end{array}\right.$

Wherein, Y represents the brightness of a pixel, U and V represents the colourity of a pixel, and R, G and B are the luminous intensity numerical value of red, green and blue in a pixel.

11. devices as described in claim 9 or 10, is characterized in that, the neighborhood territory pixel point that described first computing unit is corresponding according to any one pixel in described image, obtain the details index of any one pixel described, comprising:

Described first computing unit selects (2n+1) * (2n+1) neighborhood in described image centered by any one pixel described, arithmetic mean calculating is carried out to the brightness value of all pixels in described neighborhood, obtain the neighborhood territory pixel average brightness that any one pixel described is corresponding, wherein, n be more than or equal to 1 positive integer;

Described first computing unit by each pixel in described neighborhood square the adding up of difference of the brightness value neighborhood territory pixel average brightness corresponding with any one pixel described, obtain the details index of any one pixel described.

12. devices as described in claim 9 or 10, is characterized in that, while obtaining the low details degree of membership of each pixel described, also comprise at described second computing unit:

The details index threshold that described second computing unit is preset according to the details exponential sum of each pixel in described image, obtains the middle details degree of membership of each pixel described and high details degree of membership.

13. devices as claimed in claim 12, it is characterized in that, the details index threshold that described second computing unit is preset according to the details exponential sum of any one pixel in described image, obtains the low details degree of membership of any one pixel described, middle details degree of membership and high details degree of membership, comprising:

If the details index of any one pixel described is less than or equal to T ₁time, then the low details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the low details degree of membership of described arbitrary pixel and T ₂with the poor positive correlation of the details index of any one pixel described, with T ₂and T ₁poor negative correlation; If the details index of any one pixel described is more than or equal to T ₂time, then the low details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₁time, then the middle details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₁and be less than T ₂time, then the middle details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₁poor positive correlation, with T ₂and T ₁poor negative correlation; If the details index of a described pixel is more than or equal to T ₂and be less than or equal to T ₃time, then the middle details degree of membership of described arbitrary pixel is 1; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the middle details degree of membership of described arbitrary pixel and T ₄with the poor positive correlation of the details index of any one pixel described, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the middle details degree of membership of described arbitrary pixel is 0;

If the details index of any one pixel described is less than or equal to T ₃time, then the high details degree of membership of described arbitrary pixel is 0; If the details index of any one pixel described is greater than T ₃and be less than T ₄time, then the high details degree of membership of described arbitrary pixel and the details exponential sum T of any one pixel described ₃poor positive correlation, with T ₄and T ₃poor negative correlation; If the details index of any one pixel described is more than or equal to T ₄time, then the high details degree of membership of described arbitrary pixel is 1;

Wherein, T ₁, T ₂, T ₃and T ₄for the details index threshold preset, and meet 0 < T ₁< T ₂< T ₃< T ₄.

14. devices as described in claim 9 or 10, it is characterized in that, described graphics processing unit carries out noise reduction process to the pixel that any one low details degree of membership in described image is 1, comprising:

Described graphics processing unit filters out in (2n+1) * (2n+1) neighborhood centered by the pixel that any one low details degree of membership described is 1, maximal value Maxf, the minimum M inf of neighborhood territory pixel brightness and intermediate value Medf, wherein, $\mathrm{Maxf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Max}}\left(f(x+i,y+j)\right),\mathrm{Minf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Min}}\left(f(x+i,y+j)\right),$ $\mathrm{Medf}=\underset{-n\&le;i,j\&le;n}{\mathrm{Median}}\left(f(x+i,y+j)\right);$

Described graphics processing unit carries out medium filtering process to the pixel that any one low details degree of membership described is 1, if the brightness value that any one low details degree of membership described is the pixel of 1 is not equal to Maxf, and being not equal to Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 is constant; If the brightness value that any one low details degree of membership described is the pixel of 1 equals Maxf, or equals Minf, then any one low details degree of membership described in is that the brightness value of the pixel of 1 equals Medf.

15. devices as claimed in claim 12, is characterized in that, described graphics processing unit carries out edge enhancing process to the pixel that any one low details degree of membership in described image is less than 1, comprising:

Described graphics processing unit is less than the middle details degree of membership of the pixel of 1 and high details degree of membership according to any one low details degree of membership described, and strengthen weight coefficient for the different edge that middle details degree of membership and high details degree of membership are preset, any one low details degree of membership described is less than details degree of membership and high details degree of membership in the pixel of 1 be weighted and calculate, obtains the edge enhancing weight coefficient that any one low details degree of membership described is less than the pixel of 1;

Any one low details degree of membership described is less than the brightness value of the pixel of 1 by described graphics processing unit and Laplace sharpening template carries out convolution, obtains the skirt response value that any one low details degree of membership described is less than the pixel of 1;

Described graphics processing unit by the brightness value any one low details degree of membership described being less than the pixel of 1 and the edge that any one low details degree of membership described is less than the pixel of 1 strengthen weight coefficient be multiplied with skirt response value after results added, the pixel being less than 1 to any one low details degree of membership described carries out edge and strengthens process, obtain pixel that any one low details degree of membership described is less than 1 carry out edge enhancing after pixel.

16. devices as claimed in claim 15, it is characterized in that, the different edge enhancing weight coefficient preset for middle details degree of membership and high details degree of membership should meet following condition: the edge enhancing weight coefficient preset for high details degree of membership is greater than 0, and is less than the edge enhancing weight coefficient preset for middle details degree of membership for the edge enhancing weight coefficient that high details degree of membership is preset.