CN104794682A - Transform domain based image interpolation method and device thereof - Google Patents

Abstract

The invention provides a transform domain based image interpolation method and a device thereof. The transform domain based image interpolation method includes inputting an original image fO(m,n) to be processed, dividing the original image fO(m,n) into an intense gray area fH(m,n) and a slow gray area fL(m,n) according to difference of spectrums after the original image is transformed into the frequency domain, interpolating the intense gray area fH(m,n) and the slow gray area fL(m,n) by different methods according to different requirements for image quality, uniting images acquired after interpolation of the two areas to generate a complete image fN(m,n) which is interpolated. The transform domain based image interpolation method has the advantages of reducing calculation amount on the premise of guaranteeing quality of details and further reducing noise interference.

Description

A kind of image interpolation method based on transform domain and device

Technical field

The embodiment of the present invention relates to digital image processing techniques field, particularly relates to a kind of image interpolation method based on transform domain and device.

Background technology

In image procossing and display system, interpolation is a kind of basic technological means.To the convergent-divergent of image, rotation and frame frequency conversion, all must be realized by interpolation arithmetic.And the quality of interpolation algorithm, directly determine effect and the speed of above-mentioned process.

Existing image interpolation method, as Chinese patent, (application number: 201010292556.4) construct multiple interpolating function, solves the irretentive problem of topology of existence when prior art application carries out image interpolation based on radial basis function; Chinese patent (application number: 200910001133.X), by arranging different angle of deviation, calculates interpolation coefficient; Chinese patent (application number: 201010301367.9) for interpolation point chooses the square dot matrix of its most contiguous 2M × 2M preimage vegetarian refreshments composition, and obtain the interpolation weights coefficient of each preimage vegetarian refreshments to interpolation point according to the distance of each preimage vegetarian refreshments chosen and interpolation point; Space length information in Chinese patent (application number: 201110230492.X) computed image moving window between arbitrary interpolation point and four original image vegetarian refreshments and the spectral vector distance between original four pixels, the weights of computer memory distance and the weights of spectral vector distance, the vector value then calculating total weight value and this interpolation point realizes interpolation; Chinese patent (application number: the edge treated problem 200810028304.3) mainly for ENO interpolation method is improved; Chinese patent (application number: 200880126922.X) utilizes block-based exercise estimator to provide block-based motion vector, by this motion vector to correct image, reduces the halo effect in Interpolation Process; (application number: 201010558745.1) introduce similarity probability as weight during high-resolution pixel point estimation, and then evaluated error is minimized decreases the flaw that boundary occurs to Chinese patent.

Above-mentioned existing method respectively has advantage in interpolation quality or calculated amount, speed etc., but be all process for entire image, ignore the content character of image itself and user to the attention rate requirement of different content, be therefore difficult to take into account Disposal quality and speed, there is certain defect.

Summary of the invention

Image interpolation method based on transform domain provided by the invention and device, under the prerequisite ensureing details quality, simplify calculated amount, and can reduce noise.

The invention provides a kind of image interpolation method based on transform domain, comprising:

Input the original image f that pending size is M × N _o(m, n), m and n is described original image spatial domain coordinate variable, wherein m=0 ..., M-1, n=0 ..., N-1; ;

By described original image conversion frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n);

According to the difference to image quality requirements, to described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) adopts diverse ways to carry out interpolation respectively;

Merge described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _limage after (m, n) interpolation, generates the complete image f through interpolation _n(m, n).

Further, described by described original image conversion frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) comprising:

To described original image f _o(m, n) carries out Fourier transform to generate the frequency spectrum F of described original image _o(u, v), transformation for mula is:

$F_O(u,v)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}{f}_O(m,n)\exp (-j2\mathrm{\π}\left(\frac{\mathrm{mu}}{M}\frac{\mathrm{nv}}{N}\right)),$ Wherein u and v is described original image frequency field coordinate variable, wherein u=0 ..., M-1, v=0 ..., N-1;

By the frequency spectrum F of Hi-pass filter to described original image _o(u, v) carries out low frequency filtration, to obtain high frequency spectrum F _h(u, v);

To described high frequency spectrum F _h(u, v) carries out inverse fourier transform to generate the gray scale intense regions f in original image _h(m, n), reconstructed formula is:

$f_H(m,n)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{u=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{v=0}{\overset{N-1}{\mathrm{\Σ}}}{F}_H(u,v)\exp \left(j2\mathrm{\π}(\frac{\mathrm{mu}}{M}+\frac{\mathrm{nv}}{N})\right);$

According to described original image f _o(m, n) and described gray scale intense regions f _h(m, n) contrast is to obtain the slow region f of gray scale _l(m, n).

Further, described Hi-pass filter adopts Gauss's Hi-pass filter, then cutoff frequency is D ₀the transport function of Gauss's Hi-pass filter be:

$H(u,v)=1-e^{-D{(u,v)}^2/2D_0^2},$ Wherein $D(u,v)={[{(u-\frac{M}{2})}^2+{(v-\frac{N}{2})}^2]}^{\frac{1}{2}},$ Wherein suppose that image low-limit frequency is in the transform domain as illustrated D _l, highest frequency is D _h, then D ₀=(D _l+ D _h)/2.

Further, adopt cube method of interpolation to described gray scale intense regions, consider 16 neighborhood territory pixel points of the floating-point coordinate (i+u, j+v) of object pixel, the cube interpolation calculation formula of object pixel f (i+u, j+v) is:

f(i+u,j+v)＝[A]×[B]×[C]

Wherein, i, j are the integral part of image pixel coordinates in interpolation arithmetic, and u, v are the fraction part of image pixel coordinates in interpolation arithmetic,

[A]＝[S(u+1) S(u+0) S(u-1) S(u-2)]

$\left[B\right]=\left[\begin{array}{cccc}f(i-1,j-1)& f(i-1,j+0)& f(i-1,j+1)& f(i-1,j+2)\\ f(i+0,j-1)& f(i+0,j+0)& f(i+0,j+1)& f(i+0,j+2)\\ f(i+1,j-1)& f(i+1,j+0)& f(i+1,j+1)& f(i+1,j+2)\\ f(i+2,j-1)& f(i+2,j+0)& f(i+2,j+1)& f(i+2,j+2)\end{array}\right]$

$\left[C\right]=\left[\begin{array}{c}S(v+1)\\ S(v+0)\\ S(v-1)\\ S(v-2)\end{array}\right],$

Wherein $S\left(x\right)=\left[\begin{array}{cc}1-2\&times;\mathrm{abs}\left(x\right)^2+\mathrm{abs}\left(x\right)^3,& 0\&le;\mathrm{abs}\left(x\right)<1\\ 4-8\&times;\mathrm{abs}\left(x\right)+5\&times;\mathrm{abs}\left(x\right)^2-\mathrm{abs}\left(x\right)^3,& 1\&le;\mathrm{abs}\left(x\right)<2\\ 0,& \mathrm{abs}\left(x\right)\&GreaterEqual;2\end{array}\right].$

Further, adopt most neighbor interpolation to the slow region of described gray scale, if object pixel is f (i+u, j+v), then its computing formula is:

$f(i+u,j+v)=\left\{\begin{array}{cc}f(i,j),& u<0.5,v<0.5\\ f(i+1,j),& u\&GreaterEqual;0.5,v<0.5\\ f(i,j+1),& u<0.5,v\&GreaterEqual;0.5\\ f(i+1,j+1),& u\&GreaterEqual;0.5,v\&GreaterEqual;0.5\end{array}\right.$ Wherein, i, j are the integral part of image pixel coordinates in interpolation arithmetic, and u, v are the fraction part of image pixel coordinates in interpolation arithmetic.

The present invention also provides a kind of image interpolation device based on transform domain, comprising:

Load module, for inputting the original image f that pending size is M × N _o(m, n), m and n is described original image spatial domain coordinate variable, wherein m=0 ..., M-1, n=0 ..., N-1;

Segmentation module, for described original image is converted frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n);

Interpolation processing module, for according to the difference to image quality requirements, to described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) adopts diverse ways to carry out interpolation respectively;

Merge module, for merging described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _limage after (m, n) interpolation, generates the complete image f through interpolation _n(m, n).

Further, described segmentation module comprises:

Fourier transformation unit, for described original image f _o(m, n) carries out Fourier transform to generate the frequency spectrum F of described original image _o(u, v), transformation for mula is:

$F_O(u,v)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{m=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{f}_O(m,n)\exp (-j2\mathrm{\&pi;}\left(\frac{\mathrm{mu}}{M}\frac{\mathrm{nv}}{N}\right)),$ Wherein u and v is described original image frequency field coordinate variable, wherein u=0 ..., M-1, v=0 ..., N-1;

Low frequency filter element, for by the frequency spectrum F of Hi-pass filter to described original image _o(u, v) carries out low frequency filtration, to obtain high frequency spectrum F _h(u, v);

Inverse Fourier transform unit, for described high frequency spectrum F _h(u, v) carries out inverse fourier transform to generate the gray scale intense regions f in original image _h(m, n), reconstructed formula is:

$f_H(m,n)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{u=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{F}_H(u,v)\exp \left(j2\mathrm{\&pi;}(\frac{\mathrm{mu}}{M}+\frac{\mathrm{nv}}{N})\right);$

Contrast unit, for according to described original image f _o(m, n) and described gray scale intense regions f _h(m, n) contrast is to obtain the slow region f of gray scale _l(m, n).

Image interpolation method based on transform domain provided by the invention and device, first by carrying out fineness classification to pending picture material, and then adopt for different regions the interpolation method that quality is different from speed emphasis, under the prerequisite ensureing details quality, simplify calculated amount, and can noise be reduced.

Accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.

Fig. 1 is the process flow diagram of the image interpolation method embodiment that the present invention is based on transform domain;

Fig. 2 is the process flow diagram of image frequency domain division methods in Fig. 1;

Fig. 3 is the structural representation of the image interpolation device embodiment that the present invention is based on transform domain.

Embodiment

For making the object of the embodiment of the present invention, technical scheme and advantage clearly, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

Fig. 1 is the process flow diagram of the image interpolation method embodiment that the present invention is based on transform domain, as shown in Figure 1, specifically comprises:

Step 10, input pending original image;

Suppose that pending original image size is the original image f of M × N _o(m, n), m and n is described original image spatial domain coordinate variable, wherein m=0 ..., M-1, n=0 ..., N-1;

The frequency of image is the index of grey scale change severe degree in token image, is the gradient of half-tone information on plane space.The content of image is made up of several connected domains, and the grey scale change of each connected domain inside is slow, and corresponding frequency values is lower; And marginal portion grey scale change between each connected domain is violent, corresponding frequency values is higher.

Step 20, by described original image conversion frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n);

The gray scale intense regions f of indication in this step _hpicture frequency in (m, n) is greater than the slow region f of gray scale _l(m, n).

Step 30, the difference of basis to image quality requirements, to described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) adopts diverse ways to carry out interpolation respectively;

Typically, there are different quality and rate request in the region that user is different to frequency, therefore can choose different interpolation methods for zones of different and carry out interpolation processing respectively.

Step 40, merge described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _limage after (m, n) interpolation, generates the complete image f through interpolation _n(m, n).

The image interpolation method based on transform domain that above-described embodiment provides is first by carrying out fineness classification to pending picture material, and then adopt for different regions the interpolation method that quality is different from speed emphasis, under the prerequisite ensureing details quality, simplify calculated amount, and can noise be reduced.

Preferably, in such scheme, cube method of interpolation can be adopted to described gray scale intense regions, consider the floating-point coordinate (i+u of object pixel, j+v) 16 neighborhood territory pixel points, the cube interpolation calculation formula of object pixel f (i+u, j+v) is:

f(i+u,j+v)＝[A]×[B]×[C]

Wherein, i, j are the integral part of image pixel coordinates in interpolation arithmetic, and u, v are the fraction part of image pixel coordinates in interpolation arithmetic,

[A]＝[S(u+1) S(u+0) S(u-1) S(u-2)]

$\left[B\right]=\left[\begin{array}{cccc}f(i-1,j-1)& f(i-1,j+0)& f(i-1,j+1)& f(i-1,j+2)\\ f(i+0,j-1)& f(i+0,j+0)& f(i+0,j+1)& f(i+0,j+2)\\ f(i+1,j-1)& f(i+1,j+0)& f(i+1,j+1)& f(i+1,j+2)\\ f(i+2,j-1)& f(i+2,j+0)& f(i+2,j+1)& f(i+2,j+2)\end{array}\right]$

$\left[C\right]=\left[\begin{array}{c}S(v+1)\\ S(v+0)\\ S(v-1)\\ S(v-2)\end{array}\right],$

Wherein $S\left(x\right)=\left[\begin{array}{cc}1-2\&times;\mathrm{abs}\left(x\right)^2+\mathrm{abs}\left(x\right)^3,& 0\&le;\mathrm{abs}\left(x\right)<1\\ 4-8\&times;\mathrm{abs}\left(x\right)+5\&times;\mathrm{abs}\left(x\right)^2-\mathrm{abs}\left(x\right)^3,& 1\&le;\mathrm{abs}\left(x\right)<2\\ 0,& \mathrm{abs}\left(x\right)\&GreaterEqual;2\end{array}\right].$

This programme preferably adopts cube interpolation for gray scale acute variation region, and its technique effect is as follows:

(1) impact of the direct adjoint point gray-scale value of object pixel four is not only considered, also consider the impact of each direct adjoint point gray-value variation rate, the gray-scale value that make use of pixel in 16 neighborhoods around to be sampled does cubic interpolation calculating, and therefore computational accuracy is high, and interpolation is good;

(2) this interpolation arithmetic only carries out for gray scale acute variation region, effectively reduces the process range of high precision computation, has taken into account interpolation quality and bulk treatment speed.

Preferably, in such scheme, adopt most neighbor interpolation to the slow region of described gray scale, if object pixel is f (i+u, j+v), then its computing formula is

$f(i+u,j+v)=\left\{\begin{array}{cc}f(i,j),& u<0.5,v<0.5\\ f(i+1,j),& u\&GreaterEqual;0.5,v<0.5\\ f(i,j+1),& u<0.5,v\&GreaterEqual;0.5\\ f(i+1,j+1),& u\&GreaterEqual;0.5,v\&GreaterEqual;0.5\end{array}\right.$ Wherein, i, j are the integral part of image pixel coordinates in interpolation arithmetic, and u, v are the fraction part of image pixel coordinates in interpolation arithmetic.

The present invention is directed to the slow region of variation of gray scale and preferably adopt most neighbor interpolation, its technique effect is as follows:

(1) the method selects image information the most adjacent in four neighborhoods as the information of interpolating pixel, only with the gray-scale value of (namely the most contiguous) pixel had the greatest impact to this sampled point numerical value as this point, without the need to addition subtraction multiplication and division, therefore calculate simple, fastest;

(2) this Fast Interpolation computing is only carried out for the slow region of variation of gray scale, has taken into account interpolation quality and bulk treatment speed.

Further, Fig. 2 is the process flow diagram of image frequency domain division methods in Fig. 1, as shown in Figure 2, on the basis of technique scheme, particularly, described by described original image conversion frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) comprising:

Step 201, to described original image f _o(m, n) carries out Fourier transform to generate the frequency spectrum F of described original image _o(u, v), transformation for mula is

$F_O(u,v)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{m=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{f}_O(m,n)\exp (-j2\mathrm{\&pi;}\left(\frac{\mathrm{mu}}{M}\frac{\mathrm{nv}}{N}\right)),$ Wherein u and v is described original image frequency field coordinate variable, wherein u=0 ..., M-1, v=0 ..., N-1;

Step 202, by the frequency spectrum F of Hi-pass filter to described original image _o(u, v) carries out low frequency filtration, to obtain high frequency spectrum F _h(u, v);

In this step, preferably, described Hi-pass filter can adopt Gauss's Hi-pass filter, then cutoff frequency is D ₀the transport function of Gauss's Hi-pass filter be

$H(u,v)=1-e^{-D{(u,v)}^2/2D_0^2},$ Wherein $D(u,v)={[{(u-\frac{M}{2})}^2+{(v-\frac{N}{2})}^2]}^{\frac{1}{2}}.$

Wherein suppose that image low-limit frequency is in the transform domain as illustrated D _l, highest frequency is D _h, then D ₀=(D _l+ D _h)/2.

During use, by image conversion to frequency domain F _oafter (u, v), by F _o(u, v) is multiplied by H (u, v), can obtain filtered frequency spectrum.

Step 203, to described high frequency spectrum F _h(u, v) carries out inverse fourier transform to generate the gray scale intense regions f in original image _h(m, n), reconstructed formula is

$f_H(m,n)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{u=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{F}_H(u,v)\exp \left(j2\mathrm{\&pi;}(\frac{\mathrm{mu}}{M}+\frac{\mathrm{nv}}{N})\right);$

Step 204, according to described original image f _o(m, n) and described gray scale intense regions f _h(m, n) contrast is to obtain the slow region f of gray scale _l(m, n).

In such scheme, through the frequency spectrum of high-pass filtering, owing to having filtered out the low-frequency component in original image frequency domain, the only frequency spectrum of remaining gray scale acute variation part (i.e. edge), that obtain after therefore getting back to spatial domain by inverse fourier transform is marginal portion (gray scale intense regions f _h(m, n)) image, then to contrast with original image, then can realize the slow changing unit of gray scale in original image (the slow region f of gray scale _l(m, n)) with the segmentation of acute variation part.Technique scheme makes full use of the characteristic of variation of image grayscale severe degree at frequency domain, carrying out, can improve the speed of Iamge Segmentation without the need to considering that concrete gray scale size is unified.

The present invention also provides a kind of image interpolation device based on transform domain, and Fig. 3 is the structural representation of the image interpolation device embodiment that the present invention is based on transform domain, and as shown in Figure 3, this device comprises:

Load module 1, for for inputting the original image f that pending size is M × N _o(m, n), m and n is described original image spatial domain coordinate variable, wherein m=0 ..., M-1, n=0 ..., N-1;

Segmentation module 2, for described original image is converted frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n);

Interpolation processing module 3, for according to the difference to image quality requirements, to described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) adopts diverse ways to carry out interpolation respectively;

Merge module 4, for merging described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _limage after (m, n) interpolation, generates the complete image f through interpolation _n(m, n).

Particularly, on the basis of technique scheme, described segmentation module 2 can comprise:

Fourier transformation unit 21, for described original image f _o(m, n) carries out Fourier transform to generate the frequency spectrum F of described original image _o(u, v), transformation for mula is:

$F_O(u,v)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{m=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{f}_O(m,n)\exp (-j2\mathrm{\&pi;}\left(\frac{\mathrm{mu}}{M}\frac{\mathrm{nv}}{N}\right)),$ Wherein u and v is described original image frequency field coordinate variable, wherein u=0 ..., M-1, v=0 ..., N-1;

Low frequency filter element 22, for by the frequency spectrum F of Hi-pass filter to described original image _o(u, v) carries out low frequency filtration, to obtain high frequency spectrum F _h(u, v);

Inverse Fourier transform unit 23, for described high frequency spectrum F _h(u, v) carries out inverse fourier transform to generate the gray scale intense regions f in original image _h(m, n), reconstructed formula is:

$f_H(m,n)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{u=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{F}_H(u,v)\exp \left(j2\mathrm{\&pi;}(\frac{\mathrm{mu}}{M}+\frac{\mathrm{nv}}{N})\right);$

Contrast unit 24, for according to described original image f _o(m, n) and described gray scale intense regions f _h(m, n) contrast is to obtain the slow region f of gray scale _l(m, n).

The device of the present embodiment, may be used for the technical scheme performing embodiment of the method shown in Fig. 1 and Fig. 2, it realizes principle and technique effect is similar, repeats no more herein.

Last it is noted that above each embodiment is only in order to illustrate technical scheme of the present invention, be not intended to limit; Although with reference to foregoing embodiments to invention has been detailed description, those of ordinary skill in the art is to be understood that: it still can be modified to the technical scheme described in foregoing embodiments, or carries out equivalent replacement to wherein some or all of technical characteristic; And these amendments or replacement, do not make the essence of appropriate technical solution depart from the scope of various embodiments of the present invention technical scheme.

Claims (7)

1. based on an image interpolation method for transform domain, it is characterized in that, comprising:

Input the original image f that pending size is M × N _o(m, n), m and n is described original image spatial domain coordinate variable, wherein m=0 ..., M-1, n=0 ..., N-1;

Described original image is transformed to frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n);

According to the difference to image quality requirements, to described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) adopts diverse ways to carry out interpolation respectively;

Merge described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _limage after (m, n) interpolation, generates the complete image f through interpolation _n(m, n).

2. the image interpolation method based on transform domain according to claim 1, is characterized in that, described by described original image conversion frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) comprising:

To described original image f _o(m, n) carries out Fourier transform to generate the frequency spectrum F of described original image _o(u, v), transformation for mula is:

$F_0(u,v)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{m=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{f}_0(m,n)\exp (-j2\mathrm{\&pi;}\left(\frac{\mathrm{mu}}{M}\frac{\mathrm{nv}}{N}\right)),$ Wherein u and v is described original image frequency field coordinate variable, wherein u=0 ..., M-1, v=0 ..., N-1;

By the frequency spectrum F of Hi-pass filter to described original image _o(u, v) carries out low frequency filtration, to obtain high frequency spectrum F _h(u, v);

To described high frequency spectrum F _h(u, v) carries out inverse fourier transform to generate the gray scale intense regions f in original image _h(m, n), reconstructed formula is:

$f_H(m,n)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{u=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{F}_H(u,v)\exp \left(j2\mathrm{\&pi;}(\frac{\mathrm{mu}}{M}+\frac{\mathrm{nv}}{N})\right);$

According to described original image f _o(m, n) and described gray scale intense regions f _h(m, n) contrast is to obtain the slow region f of gray scale _l(m, n).

3. the image interpolation method based on transform domain according to claim 2, is characterized in that, described Hi-pass filter adopts Gauss's Hi-pass filter, then cutoff frequency is D ₀the transport function of Gauss's Hi-pass filter be:

$H(u,v)=1-e^{-D{(u,v)}^2/2D_0^2},$ Wherein $D(u,v)={[{(u-\frac{M}{2})}^2+{(v-\frac{N}{2})}^2]}^{\frac{1}{2}},$

Wherein suppose that image low-limit frequency is in the transform domain as illustrated D _l, highest frequency is D _h, then D ₀=(D _l+ D _h)/2.

4. the image interpolation method based on transform domain according to any one of claims 1 to 3, it is characterized in that, cube method of interpolation is adopted to described gray scale intense regions, consider the floating-point coordinate (i+u of object pixel, j+v) 16 neighborhood territory pixel points, the cube interpolation calculation formula of object pixel f (i+u, j+v) is:

F (i+u, j+v)=[A] × [B] × [C], wherein, i, j are the integral part of image pixel coordinates in interpolation arithmetic, and u, v are the fraction part of image pixel coordinates in interpolation arithmetic,

[A]＝[S(u+1) S(u+0) S(u-1) S(u-2)]

$\left[B\right]=\left[\begin{array}{cccc}f(i-1,j-1)& f(i-1,j+0)& f(i-1,j+1)& f(i-1,j+2)\\ f(i+0,j-1)& f(i+0,j+0)& f(i+0,j+1)& f(i+0,j+2)\\ f(i+1,j-1)& f(i+1,j+0)& f(i+1,j+1)& f(i+1,j+2)\\ f(i+2,j-1)& f(i+2,j+0)& f(i+2,j+1)& f(i+2,j+2)\end{array}\right]$

$\left[C\right]=\left[\begin{array}{c}S(v+1)\\ S(v+0)\\ S(v-1)\\ S(v-2)\end{array}\right],$

Wherein $S\left(x\right)=\left[\begin{array}{cc}1-2\&times;\mathrm{abs}\left(x\right)^2+\mathrm{abs}\left(x\right)^3,& 0\&le;\mathrm{abs}\left(x\right)<1\\ 4-8\&times;\mathrm{abs}\left(x\right)+5\&times;\mathrm{abs}\left(x\right)^2-\mathrm{abs}\left(x\right)^3& 1\&le;\mathrm{abs}\left(x\right)<2\\ 0,& \mathrm{abs}\left(x\right)\&GreaterEqual;2\end{array}\right].$

5. the image interpolation method based on transform domain according to any one of claims 1 to 3, is characterized in that, adopt most neighbor interpolation to the slow region of described gray scale, if object pixel is f (i+u, j+v), then its computing formula is:

$f(i+u,j+v)=\left\{\begin{array}{cc}f(i,j)& u<0.5,v<0.5\\ f(i+1,j)& u\&GreaterEqual;0.5,v<0.5\\ f(i,j+1)& u<0.5,v\&GreaterEqual;0.5\\ f(i+1,j+1)& u\&GreaterEqual;0.5,v\&GreaterEqual;0.5\end{array}\right.$ Wherein, i, j are the integral part of image pixel coordinates in interpolation arithmetic, and u, v are the fraction part of image pixel coordinates in interpolation arithmetic.

6., based on an image interpolation device for transform domain, it is characterized in that, comprising:

Load module, for inputting the original image f that pending size is M × N _o(m, n), m and n is described original image spatial domain coordinate variable, wherein m=0 ..., M-1, n=0 ..., N-1;

Segmentation module, for described original image is converted frequency domain, and different from described original image f according to frequency spectrum _o(m, n) is divided into gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n);

Interpolation processing module, for according to the difference to image quality requirements, to described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _l(m, n) adopts diverse ways to carry out interpolation respectively;

Merge module, for merging described gray scale intense regions f _hthe slow region f of (m, n) and gray scale _limage after (m, n) interpolation, generates the complete image f through interpolation _n(m, n).

7. the image interpolation device based on transform domain according to claim 6, is characterized in that, described segmentation module comprises:

Fourier transformation unit, for described original image f _o(m, n) carries out Fourier transform to generate the frequency spectrum F of described original image _o(u, v), transformation for mula is:

$F_0(u,v)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{m=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{f}_0(m,n)\exp (-j2\mathrm{\&pi;}\left(\frac{\mathrm{mu}}{M}\frac{\mathrm{nv}}{N}\right)),$ Wherein u and v is described original image frequency field coordinate variable, wherein u=0 ..., M-1, v=0 ..., N-1;

Low frequency filter element, for by the frequency spectrum F of Hi-pass filter to described original image _o(u, v) carries out low frequency filtration, to obtain high frequency spectrum F _h(u, v);

Inverse Fourier transform unit, for described high frequency spectrum F _h(u, v) carries out inverse fourier transform to generate the gray scale intense regions f in original image _h(m, n), reconstructed formula is:

$f_H(m,n)=\frac{1}{\sqrt{M}\sqrt{N}}\underset{u=0}{\overset{M-1}{\mathrm{\&Sigma;}}}\underset{v=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{F}_H(u,v)\exp \left(j2\mathrm{\&pi;}(\frac{\mathrm{mu}}{M}+\frac{\mathrm{nv}}{N})\right);$

Contrast unit, for according to described original image f _o(m, n) and described gray scale intense regions f _h(m, n) contrast is to obtain the slow region f of gray scale _l(m, n).