CN105631814A - Low pass filtering-based length exposure wide dynamic feathering fusion method - Google Patents

Abstract

The invention discloses a low pass filtering-based length exposure wide dynamic feathering fusion method. The method comprises steps: a highlight area in a long exposure image is extracted through a brightness area adjustment operator; a mean value is solved according to results obtained in the first step, the mean value is used for binary processing, and a binary mask formed by 0 and 1 is obtained; low pass filtering is carried out on the binary mask, feathering effects are generated, and a feathering mask is obtained; short exposure information and long exposure information are extracted: (1) multiplication operation is carried out on the feathering mask obtained in the third step and the short exposure image; (2) the feathering mask image obtained in the third step is subjected to negation, multiplication operation with the long exposure image is carried out; addition operation is carried out on the result obtained in the fourth step; and data correction is carried out on the result of the fifth step and the data are outputted. The method is applicable to FPGA realization, the realization is simple, the occupied hardware resources are few, the fusion method directly fuses the original pixel, and the real-time performance is high.

Description

A kind of length based on low-pass filtering exposes width and dynamically sprouts wings fusion method

Technical field

The present invention relates to a kind of wide dynamic fusion algorithm, particularly relate to a kind of length based on low-pass filtering and expose width and dynamically sprout wings fusion method.

Background technology

In various wide dynamic techniques, the wide dynamic technique merged based on multiframe is comparatively general one, and blending algorithm is also varied, such as methods such as the fusion of pyramid, segment fusions, comparatively speaking, these algorithms realize complexity, take hardware resource more, real-time is poor, is unfavorable for that product realizes.

Summary of the invention

It is an object of the invention to overcome the deficiencies in the prior art, it is provided that a kind of under the premise as far as possible not changing original image pixels value, length exposure image is directly merged, is suitable for the hard-wired length based on low-pass filtering and exposes width and dynamically sprout wings fusion method.

It is an object of the invention to be achieved through the following technical solutions: a kind of length based on low-pass filtering exposes width and dynamically sprouts wings fusion method, comprises the following steps:

S1: extract highlight regions, adjusts operator by luminance area and the highlight regions in long exposure image is extracted;

S2: set up binaryzation mask, the result that step S1 is obtained is averaged, and utilizes average to carry out binary conversion treatment, obtains a binaryzation mask being made up of 0 and 1;

S3: mask is sprouted wings, and binaryzation mask is carried out low-pass filtering so that it is produce feather effect, obtain emergence mask L " (x, y);

S4: extract short exposure information and long exposure information:

(1) the emergence mask obtained by step S3 and short exposed images do multiplication operation;

(2) the emergence mask image obtained by step S3 negates, and does multiplication operation with long exposure image;

S5: length exposure information merges, and the result obtained by step S4 does add operation;

S6: the result of step S5 is carried out data correction, then exports.

Described extraction highlight regions method is: first by each pixel F of long exposure image F (x, y) is normalized, then bring into formula L (x, y)=1/ (1+F (and x, y)^-10) obtain high luminance area area image L.

Described set up binaryzation mask method be: first ask in step S1 the acquiring method of the average M, average M of high luminance area area image L for be added by all pixels, and divided by the total pixel number of image;

Traversal high luminance area area image L, to each pixel L, (x, y) proceeds as follows: as L, (x, when y) value is be more than or equal to M, L ' (x, y)=1, as L, (x, when y) value is less than M, L ' (x, y)=0, it may be assumed that

$L^{\&prime;}(x,y)=\left\{\begin{array}{cc}1& L(x,y)\&GreaterEqual;M\\ 0& L(x,y)<M\end{array}\right.$

Wherein, M is the average of high luminance area area image L, and L ' is the binaryzation mask obtained.

The method of described extraction short exposure information and long exposure information is:

(1) the emergence mask L that step S3 obtained " (x, y) does multiplication operation, obtains a short exposure frame B ' with short exposed images B, it may be assumed that

B ' (x, y)=L " (x, y) B (x, y)

Wherein, B ' is short exposure frame, and B is short exposed images;

(2) the emergence mask image obtained by step S3 negates, and does multiplication operation with long exposure image F, obtains a long exposure information image F ', it may be assumed that

F ' (x, y)=F (x, y) [1-L " (x, y)]

Wherein, F ' is long exposure information image, and F is long exposure image.

The method that described length exposure information merges is: the long exposure information image F ' obtained by S4 and short exposure frame B ' carries out being added and merges, the image O after being merged, and in order to regulate syncretizing effect, is multiplied by a factor sigma and does effect adjustment, it may be assumed that

O (x, y)=�� B ' (x, y)+(1-��) F ' (x, y)

Wherein, �� span is [0,1], and O is the image after merging.

The method of described data correction is: traversing graph as O each pixel O (x, y), it is judged that ((x, time y) more than 255, is set to 255 by the value of data to O as O for x, value size y); When O (x, when value y) is less than 0, is set to 0 by the value of data, it may be assumed that

$O(x,y)=\left\{\begin{array}{cc}255& O(x,y)>255\\ 0& O(x,y)<0\end{array}\right..$

The invention has the beneficial effects as follows: comparing the blending algorithms such as the tradition fusion of pyramid, segment fusion, this blending algorithm realizes simple, takies hardware resource less, and image pixel is directly merged by this fusion method, and real-time is significantly high.

(1) this fusion method is except mask emergence needs employing convolution realization, and other steps are all based on the operation of pixel, is suitable for FPGA and realizes, it is achieved simply;

(2) this fusion method is based on the fusion of pixel, is directly merged by length exposure information, and the image after therefore merging has been maximally maintained artwork information, will not produce unpredictable problem.

Accompanying drawing explanation

Fig. 1 is fusion method flow chart of the present invention.

Detailed description of the invention

Below in conjunction with accompanying drawing, technical scheme is described in further detail, but protection scope of the present invention is not limited to the following stated.

The fusion method as it is shown in figure 1, a kind of length exposure width based on low-pass filtering is dynamically sprouted wings, comprises the following steps:

S1: extract highlight regions, adjusts operator by luminance area and the highlight regions in long exposure image is extracted;

Described extraction highlight regions method is: first by each pixel F of long exposure image F (x, y) is normalized, then bring into formula L (x, y)=1/ (1+F (and x, y)^-10) obtain high luminance area area image L.

S2: set up binaryzation mask, the result that step S1 is obtained is averaged, and utilizes average to carry out binary conversion treatment, obtains a binaryzation mask being made up of 0 and 1;

Described set up binaryzation mask method be: first ask in step S1 the acquiring method of the average M, average M of high luminance area area image L for be added by all pixels, and divided by the total pixel number of image;

Traversal high luminance area area image L, to each pixel L, (x, y) proceeds as follows: as L, (x, when y) value is be more than or equal to M, L ' (x, y)=1, as L, (x, when y) value is less than M, L ' (x, y)=0, it may be assumed that

$L^{\&prime;}(x,y)=\left\{\begin{array}{cc}1& L(x,y)\&GreaterEqual;M\\ 0& L(x,y)<M\end{array}\right.$

Wherein, M is the average of high luminance area area image L, and L ' is the binaryzation mask obtained.

S3: mask is sprouted wings, and binaryzation mask is carried out low-pass filtering so that it is produce feather effect, obtain emergence mask L " (x, y);

Described set up the method that binaryzation mask sprouts wings and be: the binaryzation mask that obtains in S2 is adopted known gaussian filtering or mean filter method so that it is produce feather effect, and obtain an emergence mask L " (x, y).

S4: extract short exposure information and long exposure information:

(1) the emergence mask L that step S3 obtained " (x, y) does multiplication operation, obtains a short exposure frame B ' with short exposed images B, it may be assumed that

B ' (x, y)=L " (x, y) B (x, y)

Wherein, B ' is short exposure frame, and B is short exposed images;

(2) the emergence mask image obtained by step S3 negates, and does multiplication operation with long exposure image F, obtains a long exposure information image F ', it may be assumed that

F ' (x, y)=F (x, y) [1-L " (x, y)]

Wherein, F ' is long exposure information image, and F is long exposure image.

S5: length exposure information merges, and the result obtained by step S4 does add operation;

The method that described length exposure information merges is: the long exposure information image F ' obtained by S4 and short exposure frame B ' carries out being added and merges, the image O after being merged, and in order to regulate syncretizing effect, is multiplied by a factor sigma and does effect adjustment, it may be assumed that

O (x, y)=�� B ' (x, y)+(1-��) F ' (x, y)

Wherein, �� span is [0,1], and O is the image after merging.

S6: the result of step S5 is carried out data correction, then exports;

The method of described data correction is: traversing graph as O each pixel O (x, y), it is judged that ((x, time y) more than 255, is set to 255 by the value of data to O as O for x, value size y); When O (x, when value y) is less than 0, is set to 0 by the value of data, it may be assumed that

$O(x,y)=\left\{\begin{array}{cc}255& O(x,y)>255\\ 0& O(x,y)<0\end{array}\right..$

The technical issues that need to address of the present invention have:

(1) useful information respective in long exposure image and short exposed images is split;

(2) emergence mask how is set up;

(3) emergence mask how is utilized to be split two width image corresponding regions;

(4) how two parts useful information is synthesized one without the image splicing vestige.

The above is only the preferred embodiment of the present invention, it is to be understood that the present invention is not limited to form disclosed herein, it is not to be taken as the eliminating to other embodiments, and can be used for other combinations various, amendment and environment, and in contemplated scope described herein, can be modified by the technology of above-mentioned instruction or association area or knowledge. And the change that those skilled in the art carry out and change are without departing from the spirit and scope of the present invention, then all should in the protection domain of claims of the present invention.

Claims (7)

1. the length based on low-pass filtering exposes width and dynamically sprouts wings fusion method, it is characterised in that comprise the following steps:

S1: extract highlight regions, adjusts operator by luminance area and the highlight regions in long exposure image is extracted;

S2: set up binaryzation mask, the result that step S1 is obtained is averaged, and utilizes average to carry out binary conversion treatment, obtains a binaryzation mask being made up of 0 and 1;

S3: mask is sprouted wings, and binaryzation mask is carried out low-pass filtering so that it is produce feather effect, obtain emergence mask;

S4: extract short exposure information and long exposure information:

(1) the emergence mask obtained by step S3 and short exposed images do multiplication operation;

(2) the emergence mask image obtained by step S3 negates, and does multiplication operation with long exposure image;

S5: length exposure information merges, and the result obtained by step S4 does add operation;

S6: the result of step S5 is carried out data correction, then exports.

2. a kind of length based on low-pass filtering according to claim 1 exposes width and dynamically sprouts wings fusion method, it is characterized in that: described extraction highlight regions method is: first by each pixel F (x of long exposure image F, y) it is normalized, then formula L (x is brought into, y)=1/ (1+F (x, y)^-10) obtain high luminance area area image L.

3. a kind of length based on low-pass filtering according to claim 1 exposes width and dynamically sprouts wings fusion method, it is characterized in that: described set up binaryzation mask method be: first ask for the average M of high luminance area area image L in step S1, the acquiring method of average M is for be added all pixels, and divided by the total pixel number of image;

Traversal high luminance area area image L, to each pixel L, (x, y) proceeds as follows: as L, (x, when y) value is be more than or equal to M, L ' (x, y)=1, as L, (x, when y) value is less than M, L ' (x, y)=0, it may be assumed that

$L^{\&prime;}(x,y)=\left\{\begin{array}{cc}1& L(x,y)\&GreaterEqual;M\\ 0& L(x,y)<M\end{array}\right.$

Wherein, M is the average of high luminance area area image L, and L ' is the binaryzation mask obtained.

4. a kind of length based on low-pass filtering according to claim 1 exposes width and dynamically sprouts wings fusion method, it is characterised in that: described low-pass filtering adopts gaussian filtering or mean filter.

5. a kind of length based on low-pass filtering according to claim 1 exposes width and dynamically sprouts wings fusion method, it is characterised in that: the method for described extraction short exposure information and long exposure information is:

(1) the emergence mask L that step S3 obtained " (x, y) does multiplication operation, obtains a short exposure frame B ' with short exposed images B, it may be assumed that

B ' (x, y)=L " (x, y) B (x, y)

Wherein, B ' is short exposure frame, and B is short exposed images;

(2) the emergence mask image obtained by step S3 negates, and does multiplication operation with long exposure image F, obtains a long exposure information image F ', it may be assumed that

F ' (x, y)=F (x, y) [1-L " (x, y)]

Wherein, F ' is long exposure information image, and F is long exposure image.

6. a kind of length based on low-pass filtering according to claim 1 exposes width and dynamically sprouts wings fusion method, it is characterized in that: the method that described length exposure information merges is: the long exposure information image F ' obtained by S4 and short exposure frame B ' carries out being added and merges, image O after being merged, in order to regulate syncretizing effect, it is multiplied by a factor sigma and does effect adjustment, it may be assumed that

O (x, y)=�� B ' (x, y)+(1-��) F ' (x, y)

Wherein, �� span is [0,1], and O is the image after merging.

7. a kind of length based on low-pass filtering according to claim 1 exposes width and dynamically sprouts wings fusion method, it is characterized in that: the method for described data correction is: traversing graph is as each pixel O (x of O, y), judge O (x, y) value size, as O, (x, time y) more than 255, is set to 255 by the value of data; When O (x, when value y) is less than 0, is set to 0 by the value of data, it may be assumed that

$O(x,y)=\left\{\begin{array}{cc}255& O(x,y)>255\\ 0& O(x,y)<0\end{array}\right..$