CN1632829A - A method to achieve image scaling - Google Patents

Abstract

This invention discloses an image zooming method, which comprises the following steps: the original image is of Bayer data form and the aim image is of RGB data form; to zoom each picture element in the original images to get the image in Bayer data form; then to convert the zoom image into RGB data form through plug operation to get the aim image. This method is first to zoom the image in Bayer data form till suitable size and then to convert the Bayer data form into RGB data form.

Description

A method to achieve image scaling

technical field

The invention relates to digital image processing technology, especially a method for realizing image scaling.

Background technique

The images in the real world are mostly analog images, but computers can only process digital information, so the analog images must be converted into a form suitable for computer representation before they can be processed by the computer. This process of converting analog images into digital images is done by The image acquisition module is completed. The image acquisition module samples the image, discretizes the analog image into pixels, and uses a certain color model to represent the color value of the pixels. The commonly used image acquisition module is a Bayer sensor, and the Bayer sensor uses a red-green-blue (RGB) color model to represent the color value of a pixel. The RGB color model has three color types: red (R), green (G), and blue (B). The principle of the Bayer sensor is: each pixel only collects one color type. color collection rules. Since the arrangement of pixel color types in the initial row is different in the coordinate system shown in Fig. 1, there are four different image acquisition formats as shown in Fig. 1(a) to Fig. 1(d). Among them, the pixel color type of the initial row in Figure 1(a) is collected repeatedly following the law of GR; the pixel color type of the initial line in Figure 1(b) is collected repeatedly following the law of RG; Figure 1(c) the pixel point of the initial row The color type is collected repeatedly according to the law of BG; the pixel color type of the initial row in Figure 1(d) is collected repeatedly according to the law of GB. The rows refer to the pixels arranged along the x-axis in FIG. 1( a ) to FIG. 1( d ).

In the following description, the four image data formats shown in FIG. 1 are collectively referred to as Bayer data formats. The characteristic of the Bayer data format is that each pixel has only one color type among the three color types of R, G, and B.

When using a digital camera or a computer to display images, due to the different sizes of the display screens, it is often necessary to perform a scaling operation on the collected image data. The so-called scaling refers to enlarging or reducing the image by means of interpolation. The prior art uses the process shown in FIG. 2 to scale the Bayer data format into an RGB data format that meets the size requirement. The RGB data format means that each pixel in the image has three color types of R, G, and B. Here, the image collected by the Bayer sensor is called the original image, and the original image is in the Bayer data format; the image obtained after scaling and data format conversion is called the target image, and the target image is in the RGB data format. The process of obtaining the target image from the original image will be described in detail below with reference to FIG. 2 .

Step 201: Convert the image in Bayer data format into RGB data format.

The original image collected by the Bayer sensor is in the Bayer data format, which follows the law shown in Figure 1, and each pixel in the image has only one color type. In the process of converting the Bayer data format into the RGB data format, it is necessary to use an interpolation method to make each pixel point have three color types of R, G, and B.

The following takes pixel A marked in Fig. 1(a) as an example to illustrate the process of converting Bayer data format into RGB data format. It can be seen from Figure 1(a) that pixel A is of G color type, assuming the color value is G0. The two adjacent points of point A along the X-axis direction are of R color type, assuming that the color values of these two points are R1 and R2 respectively, then the color value R0 of the R color type of pixel point A can be expressed by the formula (1) The interpolation method obtains:

R0＝0.5*(R1+R2) (1)

Similarly, the two adjacent points of point A along the Y axis are of B color type, assuming that the color values of these two points are B1 and B2 respectively, then the color value B0 of the B color type of pixel point A can be obtained by the formula (2 ) shows the interpolation method to obtain:

B0＝0.5*(B1+B2) B0＝0.5*(B1+B2) (2)

The interpolation method used in formula (1) and formula (2) is the simplest interpolation method. In practical applications, other interpolation methods can be used as needed to obtain better interpolation effects.

After the conversion of the above process, the pixel point A has three color types, and its color values are G=G0, R=R0, and B=B0. After the conversion of the above process is performed on all the pixels in the original image in the Bayer data format, the image data in the RGB data format is obtained.

Step 202: performing a scaling operation on the image data in the RGB data format to obtain a scaled RGB image.

In this step, the zooming operation on the image obtained in step 201 needs to calculate the color values of all pixels in the zoomed image, and the specific steps are as follows:

First, find the corresponding point of the pixel in the image after zooming in the image before zooming.

Secondly, according to the found corresponding points, the interpolation data points are selected in the image before zooming, and the color values of the selected interpolation data points are interpolated to obtain the color values of the corresponding pixel points in the zoomed image.

Each pixel in an image in the RGB data format has three color types. The following uses the R color type as an example to illustrate the scaling operation process.

Assumption: the size of the image before scaling is M×N. Since the images mentioned in this article are all discrete images, the size of the image refers to the number of pixels. The color value of the R type of the image before scaling is f(i, j), where 0≤i<M, 0≤j<N. The size of the zoomed image is M′×N′, and the R-type color value of the zoomed image is f′(i′j′), where 0≤i′<M′, 0≤j′<N′.

The calculation process of the R type color value f'(i', j') of the pixel point (i', j') is illustrated below with reference to Fig. 3, specifically as follows:

将坐标值为(i，j)的点作为像素点(i′，j′)在缩放前图像的参考点。如图3所示，对应点(x，y)距参考点(i，j)的水平和垂直距离分别为dx和dy，则dx＝x-i，dy＝y-j，且0≤dx≤1，0 ≤dy≤1。Let x=i'M/M', y=j'N/N'. Here, the point whose coordinate value is (x, y) is called the corresponding point of the zoomed pixel point (i′, j′) in the image before zooming. From the value of (x, y) can be obtained:

The point whose coordinate value is (i, j) is used as the reference point of the image of the pixel point (i′, j′) before scaling. As shown in Figure 3, the horizontal and vertical distances from the corresponding point (x, y) to the reference point (i, j) are dx and dy respectively, then dx=xi, dy=yj, and 0≤dx≤1, 0≤ dy≤1.

After the coordinate values of the corresponding point (x, y) and the reference point (i, j) are calculated, an interpolation data point can be selected according to the corresponding point (x, y) or the reference point (i, j). For different interpolation methods, the number and position of selected interpolation data points are different.

If the bilinear interpolation method (Bi-Linear) is used to calculate the color value f'(i', j') of point (i', j'), the four points A, B, C, and D marked in Figure 3 can be selected The color value is interpolated, and the result of the interpolation can be obtained from the formula (3):

f(i,j)=(1-dx)(1-dy)f(i,j)+dy(1-dx)f(i,j+1)+dx(1-dy)f(i+1 ,j)+dxdyf(i+1,j+1)(3)

If the bicubic spline interpolation method (Bi-Cubic) is used to calculate the color value f'(i', j') of point (i', j'), the color of the sixteen points shown in gray in Figure 3 can be selected Values are interpolated.

All the M'×N' pixels in the scaled image can be calculated using the above method to obtain the R-type color values of all pixels.

A similar method can be used to calculate the B-type color value and the G-type color value. It is only necessary to substitute the color value f(i, j) in the formula (3) with the color value of the corresponding type, which will not be repeated here.

It can be seen from step 202 that the scaling method used in the prior art needs to perform an interpolation calculation on the three color types of each pixel to completely obtain the color value of the pixel, which means that each pixel needs Perform cubic interpolation calculations, and this processing method does not provide more image information. In practical applications, especially in occasions with high real-time requirements such as video processing, the method adopted in the prior art increases the complexity of calculation without improving the image quality, so it is unreasonable.

Contents of the invention

In view of this, the main purpose of the present invention is to provide a method for realizing image scaling, which reduces the amount of calculations during image scaling, reduces the computational complexity of image scaling, and speeds up image processing, so that it can be better applied to video image scaling, etc. Occasions with high real-time requirements.

In order to achieve the above object, the technical solution of the present invention is specifically realized in the following way:

A kind of method that realizes image scaling, original image is Bayer data format, purpose image is RGB data format, it is characterized in that, this method comprises the following steps:

a. Perform a scaling operation on each pixel in the original image, and obtain an image in Bayer data format after scaling;

b. Convert the scaled image obtained in step a into RGB data format to obtain the target image.

Further, the zooming operation on pixels described in step a includes the following steps:

a1. Determine the color type of the pixel in the zoomed image;

a2. Select one or more interpolation data points of the same color type as the pixel points in step a1 in the image before zooming, and use the color values of the selected interpolation data points to interpolate to calculate the color value of the pixel points in step a1.

Further, the operation of determining the color type of the pixels in the zoomed image in step a1 specifically includes: first, setting the initial row arrangement of the zoomed image; secondly, according to the initial row arrangement of the zoomed image and the pixel points described in step a1 The coordinate value of determines the color type of the pixel described in step a1.

Preferably, before the step a2, it further includes: selecting the corresponding point of the pixel in the step a1 in the image before zooming;

Then, the method for selecting interpolation data points in step a2 is specifically: selecting interpolation data points according to the corresponding points.

Further, the operation of selecting the corresponding point in step a21 is specifically: according to the coordinate value of the pixel point in step a1 and the size of the image before and after zooming, the corresponding point of the pixel point in step a1 in the image before zooming is obtained.

Preferably, the operation of selecting interpolation data points in step a22 is specifically: firstly, selecting a rectangular grid area surrounding the corresponding points described in step a21 in the image before zooming; secondly, selecting all the points falling within the rectangular grid area and Points on the rectangular grid having the same color type as the pixel points described in step a1 are used as interpolation data points.

Preferably, the rectangular grid area is parallel to the coordinate axis.

Preferably, the pixel points in step a1 are of G color type, then the rectangular grid area forms an angle of 45 degrees with the coordinate axis.

It can be seen from the above technical solution that the image zooming method of the present invention first zooms the image in the Bayer data format to meet the size requirement, and then converts the Bayer data format into an RGB data format. Such a processing method makes it only necessary to perform an interpolation operation on the pixels of the zoomed image during the scaling operation, and the calculation amount is only one-third of the existing technology, thus speeding up the image processing speed and making the method more suitable for video Processing and other fields with high real-time requirements.

Description of drawings

Figure 1(a) to Figure 1(d) are images in Bayer data format collected by Bayer sensors;

Fig. 2 is the flowchart that the image with Bayer data format is zoomed into RGB data format in the prior art;

Fig. 3 is the selection schematic diagram of interpolation data point in the prior art;

Fig. 4 is the flowchart that the image with Bayer data format is zoomed into RGB data format among the present invention;

Fig. 5 is the flow chart of image scaling operation in the present invention;

Fig. 6 is a schematic diagram of selecting interpolation data points of R or B type pixel points in the image before zooming in a preferred embodiment of the present invention;

Fig. 7 is a schematic diagram of selection of interpolation data points of G type pixel points in the image before zooming in a preferred embodiment of the present invention.

Detailed ways

The core idea of the present invention is: the original image is in Bayer data format, and the target image is in RGB data format, each pixel in the original image is scaled, and the image in Bayer data format is obtained after zooming; then, the zoomed image is passed through The interpolation operation is converted into RGB data format to obtain the target image.

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 with reference to the accompanying drawings and examples.

As shown in FIG. 4 , the present invention scales the image in the Bayer data format into an RGB data format that meets the size requirement in two stages.

The first stage: the scaling operation of the image is completed, and the scaled image still maintains the Bayer data format.

In the following description, the pixel color value of the image before scaling is expressed as f(i, j), and the pixel color value of the image after scaling is expressed as f'(i', j').

Figure 5 shows the calculation process of the color value of a certain pixel point (i′, j′) in the zoomed image, which specifically includes the following steps:

Step 501: Transform the initial row arrangement of the Bayer data format, and determine the color type of the pixel (i′, j′) according to the initial row arrangement of the scaled image.

Assuming that the initial row arrangement of the image before scaling is shown in Figure 1(a), in this step, the system can flexibly choose any of the four initial row arrangements in Figure 1 (a)-(d) as the zooming method. The Bayer data format of the post image. That is: in the present invention, the initial row arrangement of the image before and after zooming may be different.

After the initial row arrangement of the scaled image is determined, the color type of the pixel point (i′, j′) is uniquely determined by the initial row arrangement and the coordinate value of (i′, j′).

After scaling the image, select the initial row arrangement shown in Figure 1(a), then the color type of the pixel point (i′, j′) is determined by the formula (4):

After zooming, the image selects the initial row arrangement shown in Figure 1(b), then the color type of the pixel point (i′, j′) is determined by the formula (5):

After scaling the image, select the initial row arrangement shown in Figure 1(c), then the color type of the pixel point (i′, j′) is determined by the formula (6):

After scaling the image, select the initial row arrangement shown in Figure 1(d), then the color type of the pixel point (i′, j′) is determined by the formula (7):

The method of the present invention can change the initial line arrangement of the image according to actual needs, bringing great flexibility to subsequent data processing.

Step 502: Select the corresponding point (x, y) or reference point (i, j) of the pixel point (i', j') in the image after scaling in the image before scaling.

The method of selecting the corresponding point is as follows: let x=i′M/M′, y=j′N/N′, then the position of the corresponding point (x, y) is shown in Fig. 6 or Fig. 7 .

The method of selecting a reference point is as follows: in the coordinate system shown in Figure 6 or Figure 7, select the upper left corner of the corresponding point (x, y), the closest distance to the point (x, y), and the pixel point (i ', j') with the same color type as the reference point of the pixel point (i', j'), and the point (i, j) in Figure 6 or Figure 7 is the selected reference point.

Steps 503-504: Select the interpolation data point according to the corresponding point (x, y) or the reference point (i, j), and calculate the color value of the pixel point (i′, j′).

From the distribution in Figure 1, it can be seen that the number of pixels of R color type and B color type in the Bayer data format is the same, and the number of pixels of G color type is twice that of R color type or B color type, that is, G The ratio of :R:B is 2:1:1, and the selection methods of interpolation data points for pixels with different color types are discussed below:

First, the selection of interpolation data points of R or B type pixels. As shown in FIG. 6 , point A is the reference point (i, j) selected in step 502 .

If the Bi-Linear interpolation method is used to calculate the color value f'(i', j') of point (i', j'), the color values of the four points A, B, C, and D indicated in Figure 6 can be selected for interpolation Calculation, the four points A, B, C, and D are located on the rectangular grid parallel to the coordinate grid shown in Figure 6. In the coordinate system of Figure 6, the coordinate values of the four points A, B, C, and D are respectively: (i, j), (i, j+2), (i+2, j) and (i+2, j+2), the result of interpolation can be obtained from formula (8):

f'(i', j')=(1-dx)(1-dy)f(i,j)+dy(1-dx)f(i,j+2)+dx(1-dy)f( i+2,j)+dxdyf(i+2,j+2)(8)

Wherein, the interpolation parameters dx and dy of the interpolation data points are calculated as follows: by step 502, x=i'M/M', y=j'N/N', then the corresponding point (x, y) is far from the reference point ( The horizontal and vertical distances of i, j) are dx' and dy' respectively, then dx'=x-i, dy'=y-j. The values obtained after normalizing dx' and dy' to the interval [0, 1.0) using formula (9) are dx and dy.

The selection of interpolation data points in the Bi-Linear interpolation method is similar to that of the prior art, except that the intervals of the interpolation data points in the present invention in the X-axis direction and the Y-axis direction are twice that of the prior art.

If the Bi-Cubic interpolation method is used to calculate the color value f'(i', j') of point (i', j'), the color values of the sixteen points shown in gray in Figure 6 can be selected for interpolation calculation.

Second, the selection of the interpolation data points of the G-type pixel points. As shown in FIG. 7 , point A is the reference point (i, j) selected in step 502 .

If the Bi-Linear interpolation method is used to calculate the color value f'(i', j') of point (i', j'), the color values of the four points A, B, C, and D indicated in Figure 7 can be selected for interpolation Calculate, the four points A, B, C, and D are located on the rectangular grid at an angle of 45 degrees to the coordinate grid shown in Figure 7. In the coordinate system of Figure 7, the coordinates of the four points A, B, C, and D The values are: (i, j), (i-1, j+1), (i+l, j+1) and (i, j+2), the interpolation result can be obtained from formula (10):

f'(i', j')=(1-dx)(1-dy)f(i,j)+dy(1-dx)f(i-1,j+1)+dx(1-dy) f(i+1，j+1)+dxdyf(i，j+2) (10)

Wherein, the interpolation parameters dx and cy of the interpolation data points are calculated as follows: by step 502, x=i'M/M', y=j'N/N', the corresponding point (x, y) is far from the reference point (i , j) the horizontal and vertical distances are dx' and dy' respectively, then dx'=x-i, dy'=y-j. The values obtained after normalizing dx' and dy' to the interval [0, 1.0) using formula (11) are dx and dy.

As can be seen from Fig. 7, this preferred embodiment of the present invention selects all points with the same color type as (i', j') on the rectangular grid that forms an angle of 45 degrees with the X, Y coordinate grid as interpolation data point. The interpolation data points selected by this 45-degree rectangular grid are the closest to the pixel point (i′, j′) on the whole, so the result of interpolation can describe the color value of the pixel point (i′j′) more accurately .

For the selection of interpolation data points of G-type pixels, the Bi-Linear interpolation method can also adopt a selection method similar to that of R or B-type pixels, that is, the rectangular grid surrounded by interpolation data points is parallel to the coordinate grid. However, for G-type pixels, this selection method ignores the data points closer to (i′, j′), so the interpolation result is not as good as the interpolation data points selected by the 45-degree rectangular grid. The result is accurate.

If the Bi-Cubic interpolation method is used to calculate the color value f'(i', j') of point (i', j'), the color values of the sixteen points shown in gray in Figure 7 can be selected for interpolation calculation.

The above method can be used to obtain the color values of all M′×N′ pixels in the image after scaling. Compared with the prior art, the method of the present invention only needs to calculate the interpolation of M'×N' pixels, while the prior art needs to calculate the interpolation of M'×N' pixels three times, so the method of the present invention greatly reduces Reduce the amount of calculation and speed up image processing.

The second stage: convert the image data in Bayer data format into RGB data format.

This step is the same as step 201 in the prior art, and will not be repeated here.

It can be seen from the above-mentioned embodiments that the method for realizing image zooming of the present invention reduces the computational complexity of the image zooming operation, and the calculation amount is only one-third of that of the prior art, thus speeding up the image processing speed and making the method more applicable It is suitable for fields with high real-time requirements such as video processing.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of the present invention. within the scope of protection.

Claims (8)

1, a kind of method that realizes image zoom, original image is the Bayer data layout, and the purpose image is the RGB data layout, it is characterized in that, and this method may further comprise the steps:

A, each pixel in the original image is carried out zoom operations, obtain the image of Bayer data layout behind the convergent-divergent;

Image transitions becomes the RGB data layout behind b, the convergent-divergent that step a is obtained, obtains the purpose image.

2, method according to claim 1 is characterized in that, the described zoom operations to pixel of step a may further comprise the steps:

A1, determine behind the convergent-divergent color type of this pixel in the image;

A2, before convergent-divergent, choose the interpolated data point that one or more and the described pixel of step a1 have the same color type in the image, utilize the color value interpolation calculation of selected interpolated data point to go out the color value of the described pixel of step a1.

3, method according to claim 2 is characterized in that, determines among the step a1 behind the convergent-divergent that the color type operation of pixel specifically comprises in the image: at first, the initial row arrangement mode of image behind the convergent-divergent is set; Secondly, according to the color type of the described pixel of coordinate figure determining step a1 of the initial row arrangement mode of image behind the convergent-divergent and the described pixel of step a1.

4, method according to claim 2 is characterized in that, further comprises before the described step a2: choose the corresponding point in the image before convergent-divergent of pixel described in the step a1;

Then the described method of choosing interpolated data point of step a2 specifically: choose the interpolated data point according to described corresponding point.

5, method according to claim 4, it is characterized in that, the operation of choosing corresponding point among the step a21 specifically: the size according to image before and after the coordinate figure of the described pixel of step a1 and the convergent-divergent obtains the described pixel of step a1 corresponding point in the image before convergent-divergent.

6, method according to claim 4 is characterized in that, the operation of choosing interpolated data point among the step a22 specifically: at first, before convergent-divergent, choose the rectangular node district that surrounds the described corresponding point of step a21 in the image; Secondly, choosing all drops on dropping in rectangular node the district in and has the point of same color type as the interpolated data point with the described pixel of step a1 on the rectangular node.

7, method according to claim 6 is characterized in that, described rectangular node district is parallel with coordinate axis.

According to claim 6 or 7 described methods, it is characterized in that 8, the described pixel of step a1 is the G color type, then described rectangular node district becomes miter angle with coordinate axis.

Images