CN106504279A - Coloured image auto focusing method - Google Patents

Abstract

The invention discloses a color image automatic focusing method, which comprises acquisition of color image antagonism map, focus sampling point acquisition based on focus window and improved Brenner's sharpness evaluation function. The method of the present invention firstly makes full use of the color information of the color image, avoiding the loss of color details caused by only using the grayscale image to evaluate the sharpness; secondly, comprehensively considering the characteristics that the main target is mostly located in the center of the field of view and the golden section The importance of points reduces the amount of calculations due to statistical color information and the negative impact of non-target areas, effectively improving the real-time and accuracy of focusing; finally, by improving the Brenner function, without increasing the amount of calculations Next, it reflects the sharpness of the whole image more comprehensively.

Description

Color Image Autofocus Method

technical field

The invention belongs to the technical field of image processing, and more particularly relates to an automatic focusing method for a color image.

Background technique

With the development of instrument automation and intelligence, image-based autofocus has become a key technology for imaging systems such as cameras, video cameras, and microscopes. The key to realizing image-based autofocus technology is to construct a fast and reliable sharpness evaluation function, and the position corresponding to the extremum of the function is the focus position of imaging. Since the image sharpness evaluation function needs to evaluate the defocus amount of a large number of images in the complete process of the imaging system gradually approaching the focus position to just at the focus position and then away from the focus position, a good image sharpness function In addition to having unbiasedness, unimodality, sharpness and noise resistance, it is also necessary to facilitate calculation as much as possible to meet the real-time requirements of the autofocus process.

In order to reduce the amount of calculation, the existing sharpness evaluation functions are only calculated for the grayscale image, resulting in the loss of a large amount of color information. However, in a large number of actual scene images such as biomedicine, the color information of the target is very important, and the loss of color information may lead to the inability to obtain the best focus position; and if the sharpness evaluation function is calculated for the RGB three color channel images, it is difficult to Avoid the problem of greatly increasing the amount of calculation.

Contents of the invention

In view of the above defects or improvement needs of the prior art, the present invention provides an automatic focusing method for color images, the purpose of which is to make full use of image color information and realize fast automatic focusing. The technical scheme that the present invention proposes is as follows:

A color image automatic focusing method is characterized in that the method comprises the following steps:

(1) input the original image of RGB format, calculate and obtain red-green antagonism figure I _rg and blue-yellow antagonism figure I _by respectively;

(2) Obtain the focused sampling point;

First divide the original image into three types of focus windows: center window, four-side window and four-corner window;

Set different sampling strategies for the red-green antagonism map I _rg and the blue-yellow antagonism map I _by in different focusing windows: the pixels in the center window are fully sampled, the pixels in the four-side windows are sampled at intervals horizontally, and the pixels in the four-corner windows are sampled at intervals. Pixels are sampled horizontally and vertically at intervals respectively; after sampling, a linked list of sampling points L _rg corresponding to the red-green antagonism graph I _rg , and a linked list of sampling points B L _by corresponding to the blue-yellow antagonism graph I _by , are respectively obtained. The pixel position information of the focused sampling point is recorded in the sampling point linked list A L _rg and the sampling point linked list B L _by ;

(3) Use the improved Brenner function to calculate the function value of the red-green antagonism map I _rg focused sampling point and the function value F _MBrenner of the blue-yellow antagonism map I _by focused sampling point, and count all function values greater than a given threshold T The sum is used to evaluate the sharpness of color images through this statistical value.

The specific calculation process of described step (1) is as follows:

I _rg = RG

I _by = BY

in,

Y=(R+G)/2

R, G, and B are the red, green, and blue color channels of the input color image, respectively.

In the step (2): the four subimages adjacent to the center point of the original image form the central window W _a ; the four subimages at the four corners of the original image form the four corner window W _c ; in the original image, the four corner window W _c and the central window W _a The parts other than constitute the four-sided window W _b .

The improved Brenner function calculation formula in described step (3) is as follows:

Among them, L _c .length is the length of the sampling point linked list L _c ,

MB _c (i)=|I _c (L _c [i].x+2, L _c [i].y)-I _c (L _c [i].x, L _c [i].y)|×

|I _c (L _c [i].x, L _c [i].y+2)-I _c (L _c [i].x, L _c [i].y) | c=rg, by

Wherein, (L _c [i].x, L _c [i].y) is the pixel position of the i-th element in the sampling point linked list L _c .

The beneficial effect that the present invention can reach is as follows:

Due to the high real-time requirements of autofocus, in the design of a large number of sharpness evaluation functions, color information is lost in order to reduce the amount of calculation, so that it is impossible to accurately evaluate the sharpness of color images in some specific occasions. However, the method proposed by the present invention maximizes the use of the color features of the image without increasing the amount of calculation, and the evaluation result of the definition of the color image is more accurate.

Description of drawings

Fig. 1 is the flowchart of the color image automatic focusing method of the present invention;

Fig. 2 is a schematic diagram of division of focus windows in the present invention.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

Step 1, calculate the red-green antagonism graph and the blue-yellow antagonism graph:

(1.1) input the original image of RGB mode, its width is counted as W, and its height is recorded as H; the original image is decomposed into three color channel images of red (R), green (G) and blue (B);

(1.2) Utilize two color channel images of red and green to obtain a yellow (Y) channel image;

Y=(R+G)/2

(1.3) Utilize R, G, B, Y to calculate and obtain red-green antagonistic figure I _rg and blue-yellow antagonistic figure I _by ;

I _rg = RG

I _by = BY

Step 2: Sampling the pixel positions of the same size as the original image, respectively constructing the sampling point linked list A L _rg and the sampling point linked list B L _by . The sampling point linked list A L _rg will be used for the calculation of the clarity of the subsequent red-green antagonism map I _rg , and the sampling point linked list B L _by will be used for the calculation of the clarity of the blue-yellow antagonism map I _by . The purpose of constructing the linked list of sampling points is to reduce the calculation amount of the subsequent sharpness evaluation function by extracting representative focused sampling points. The selection principle of the focus sampling points of the red-green antagonism map I _rg and the blue-yellow antagonism map I _by is to make the positions of the two sampling point linked lists complementary as much as possible, and make full use of the entire image as much as possible while reducing the amount of calculation color information. Specific steps include:

(2.1) Divide the input original image into sixteen sub-pictures according to the size, and the size of each sub-picture is W/4×H/4;

(2.2) As shown in Figure 2, the sixteen sub-pictures are divided into three types of focus windows: central window W _a , four-side window W _b and four-corner window W _c ;

The four subimages adjacent to the center point of the original image constitute the central window W _a ;

Four sub-pictures at the four corners of the original image constitute the four-corner window W _c ;

In the original image, the parts other than the four-corner window W _c and the central window W _a constitute the four-side window W _b .

(2.3) record the position (x, y) of each pixel in the central window W _a , add in the sampling point linked list A L _rg ; record the position of each pixel in the central window W _a , add in the sampling point linked list B L _by ;

(2.4) The pixels in the four-sided window W _b are horizontally sampled at intervals, specifically, record the positions (x, y) of the odd-numbered pixels in the four-sided window W _b line by line, and add it to the sampling point linked list A L In _rg ; record the position (x, y) of the even-numbered pixel in the four-side window W _b line by line, and add it in the sampling point linked list B L _by ;

(2.5) Perform 2 times down-sampling on the pixels in the four-corner window W _c , that is, perform horizontal and vertical sampling at intervals, and add the sampling pixel position to the sampling point linked list A L _rg or the sampling point linked list B L _by .

Specifically, the four-corner window W _c is divided into several sub-windows according to 2×2 pixels, and the position (x, y) of the upper-left corner pixel in the sub-window is added to the sampling point link list A L _rg , and the sub-window The position (x, y) of the pixel in the lower right corner of the pixel is added to the sampling point linked list B L _by .

Since the main target is mostly located in the center of the field of view, the sampling rate of the pixels in the central window W _a is the highest; at the same time, the role of the golden section point in the photographic composition is very important, so the size setting of the central window W _a makes the golden section point exactly located in it. Compared with the central window W _a , the non-target areas in the four-sided windows W _b and the four-cornered windows W _c increase, so the sampling rates of these two types of windows are lower. The focus window and sampling strategy are set so that the sum of the lengths of the two sampling point linked lists is 9/8×W×H, which is slightly larger than the original image size.

Step 3, use the improved Brenner function to count the sum of the sharpness evaluation function values F _MBrenner of all focused sampling points (x, y):

(3.1) Utilize the improved Brenner function formula to count the function value of each pixel in the red-green antagonism graph I _rg belonging to the sampling point linked list A L _rg and the blue-yellow antagonistic graph I _by belonging to the sampling point linked list B L _by The function value for each pixel in :

MB _c (i)=|I _c (L _c [i].x+2, L _c [i].y)-I _c (L _c [i].x, L _c [i].y)|×

|I _c (L _c [i].x, L _c [i].y+2)-I _c (L _c [i].x, L _c [i].y) | c=rg, by

Wherein, (L _c [i].x, L _c [i].y) is the pixel position of the i-th element in the sampling point linked list L _c .

The improved Brenner function not only considers the grayscale difference of pixels with a difference of two units in the horizontal direction, but also considers the grayscale difference of pixels with a difference of two units in the vertical direction, which can more comprehensively reflect the local gradient changes of the entire image. At the same time, Compared with the original Brenner function, the amount of calculation is not increased too much.

(3.2) Count the sum of all function values F _MBrenner greater than a given threshold T, which is the sharpness of the input color image.

Wherein, L _c .length is the length of the sampling point linked list L _c .

The corresponding program of this method can be embedded in the image acquisition equipment. During the operation process, by comparing several calculation results and selecting the highest statistical value, the optimal focusing scheme can be selected in real time.

The color image automatic focusing method provided by the present invention firstly makes full use of the color information of the color image, avoiding the loss of color details caused by only using the grayscale image to evaluate the sharpness; secondly, considering that the main target is mostly located in the center of the field of view The characteristics of the golden section point and the importance of the golden section point reduce the amount of calculations due to statistical color information and the negative impact of non-target areas, effectively improving the real-time and accuracy of focusing; finally, by improving the Brenner function, without Under the premise of increasing the amount of calculation, it can more fully reflect the clarity of the entire image.

Claims (4)

1. A method for automatic focusing of a color image, said method comprising the steps of:

(1) inputting an original image in RGB format, and respectively calculating to obtain a red-green antagonistic graph I_rgAnd blue-yellow antagonism map I_by；

(2) Acquiring a focused sampling point;

firstly, dividing an original image into three types of focusing windows: the window comprises a central window, four-side windows and four-corner windows;

antagonism of red-green within different focus windows_rgAnd blue-yellowAntagonistic Panel I_bySetting different sampling strategies: carrying out full sampling on pixels in the central window, carrying out transverse point separation sampling on pixels in the four-side window, and respectively carrying out point separation sampling on pixels in the four-corner window in the transverse direction and the longitudinal direction; after sampling, obtaining a red-green antagonistic graph I_rgCorresponding sampling point chain table A L_rgAnd blue-yellow antagonism map I_byCorresponding sampling point chain table B L_byThe sampling point chain table A L_rgSampling point chain table B L_byThe pixel position information of a focusing sampling point is recorded;

(3) separate calculation of the Red-Green antagonism map I Using the modified Brenner function_rgFunction value of focusing sampling point and blue-yellow antagonism graph I_byFunction value F of focused sampling point_MBrennerAnd counting the sum of all function values larger than a given threshold value T, and evaluating the definition of the color image through the counted value.

2. The color image sharpness evaluation method according to claim 1, wherein the step (1) is implemented by the following steps:

I_rg＝R-G

I_by＝B-Y

wherein,

Y＝(R+G)/2

r, G, B are the three color channels of the input color image, red, green, and blue, respectively.

3. A color image sharpness evaluation method according to claim 1 or 2, characterized in that: in the step (2), 4 sub-graphs adjacent to the center point of the original image form a center window W_a(ii) a 4 sub-graphs of four corners of the original image form a four-corner window W_c(ii) a In the original image, the four corner windows W_cA central window W_aThe other parts constituting a quadrilateral window W_b。

4. A color image sharpness evaluation method according to claim 1 or 2, characterized in that the modified Brenner function calculation formula in the step (3) is as follows:

$F_{MBrenner}=\underset{c=rg,by}{\mathrm{\Σ}}\underset{i=1}{\overset{L_c.length}{\mathrm{\Σ}}}\[{\mathrm{MB}}_c\left(i\right)>T\]$

wherein L is_cLength is the sampling point chain table L_cThe length of (a) of (b),

MB_c(i)＝|I_c(L_c[i].x+2，L_c[i].y)-I_c(L_c[i].x，L_c[i].y)|×|I_c(L_c[i].x，L_c[i].y+2)-I_c(L_c[i].x，L_c[i].y)| c＝rg，by

wherein (L)_c[i].x，L_c[i]Y) is a linked list L of sampling points_cThe pixel position of the ith element.