CN105243660B - A kind of quality evaluating method of the auto-focusing image containing light source scene - Google Patents

Abstract

The invention discloses a method for evaluating the quality of an auto-focus image containing a light source scene. The method uses a saturated pixel template to act on a gradient value matrix, removes the influence of saturated pixels on the auto-focus evaluation function, and obtains the method of removing saturated pixels. The evaluation function, for the light source and overbright scene in the night scene, the evaluation function for removing saturated pixels has good unbiasedness, unimodality, sensitivity, and stability.

Description

A quality assessment method for auto-focus images of scenes containing light sources

technical field

The invention relates to a method for evaluating the quality of an automatic focus image of a scene containing a light source.

Background technique

With the rapid development of digital image and multimedia technology, more and more people pay attention to the automatic focus method based on image processing. The image processing method mainly includes the following three aspects, which are: the selection of the focus window, the selection of the focus evaluation function, and the selection of the focus search strategy. This autofocus method is to use a certain digital image processing algorithm to obtain the characteristic value of image clarity, and control the motor according to this characteristic value to drive the lens or image sensor to change the position and make corresponding adjustments until the characteristic value is satisfied. to a pre-agreed condition. The key issue in this process is how to choose an appropriate focus evaluation function.

An ideal focus evaluation function should be unbiased, unimodal, sensitive, and stable. Due to the importance of the focus evaluation function, the evaluation of image sharpness has become a hot research field. At present, the common image sharpness evaluation functions mainly include energy variance method, energy gradient method, entropy function method, spectral function method, Lapla Adams energy method, etc. The most commonly used method is the evaluation method based on the gradient, which judges the degree of blurring of the image based on the principle that the edge information of the blurred image is blurred, while the edge of the clear image is sharper and the information is obviously rich in details.

For night scene shooting, the light contrast of the scene at night will be relatively strong. At the same time, there will inevitably be lights in the night scene, and sometimes our subject is just the light. However, due to the limitation of the dynamic range of the camera sensor, when the detector image When the number of photons received by the element exceeds the maximum intensity range, it will be saturated, causing the light intensity at the pixel to be truncated at the maximum output value. We call this pixel a saturated pixel. During the autofocus process, when the scene is in-focus to out-of-focus, the saturated pixels will scatter as the degree of defocus becomes larger. Taking a point light source as an example, when it is in focus, its image point should be a point image for an ideal imaging system. However, the imaged light intensity is not the real light intensity of the light source, but a saturated pixel. For the out-of-focus situation , the point light source will be diffused to a certain extent, and the diffused surrounding pixels are still saturated pixels at this time, forming a diffused circle composed of saturated pixels, which is reflected in the obtained image as a bright spot with the same brightness. It is far from the real scene, and it is impossible to accurately evaluate the blur degree only based on the existing auto-focus evaluation function. Therefore, for night scenes or lighting scenes, finding an autofocus evaluation function suitable for saturated pixels is a key issue in autofocus.

Contents of the invention

The purpose of the present invention is to propose a method for evaluating the quality of auto-focus images of scenes containing light sources, aiming at the problems existing in the auto-focus of night scenes at present.

The object of the present invention is achieved by the following technical solutions: a method for evaluating the quality of an autofocus image containing a light source scene, the method comprising the following steps:

(1) Randomly select a focus window I _c containing a light source in the auto-focus image I containing a light source scene, and use the gradient absolute value operator to calculate the gradient value at each pixel in the focus window I _c to form the gradient value of the focus window matrix G;

(2) Normalize the gray value of _Ic , and perform binarization through the threshold T to obtain the binary image M, and extract the saturated pixel area template _BIc :

BI _c ＝(I _c >T) (1)

(3) Expand the template BI _c of the saturated pixel area by means of morphological expansion to obtain the expanded area template BI _c ' of the saturated pixel, and the expanded binary image is M';

Among them, Rid is a structural element of R×R; is an expansion operation, and M is a binarized saturated pixel expansion region template;

(4) Invert the expanded binary image M' to obtain the inverted binary image and calculate the binary map , the number N of pixels with a pixel value of 1;

(5) Binary image Act on the gradient value matrix G of the focus window to obtain the gradient information matrix W at the unsaturated pixels:

(6) Calculate the sum S _w of all gradient element values of the gradient information matrix W at the unsaturated pixel, and obtain the final definition evaluation function index Metric for desaturated pixels:

Metic＝S _w /N(5)

Among them, i and j are the rows and columns in the W template respectively, W(i, j) is the gradient value at the i-th row and j-th column in the gradient information matrix W, _a and _b are the length and length of the focus window I _c width.

Further, the threshold T in the step 2 is 0.6-0.8.

Further, R in the step 3 is 3.

The beneficial effect of the present invention is that: the present invention uses the saturated pixel template to act on the gradient value matrix, removes the influence of saturated pixels on the autofocus evaluation function, and has good unbiasedness and unimodality for light sources and overbright scenes in night scenes , sensitivity, stability.

Description of drawings

Fig. 1 is the flowchart of inventive method;

Fig. 2 is a schematic diagram of a group of real-shot out-of-focus sequences of 5 optional frames;

FIG. 3 is a real-shot scene diagram at a quasi-focus position;

Fig. 4 is the focus window area of the real shot scene graph at the quasi-focus position;

FIG. 5 is a schematic diagram of a saturated pixel area;

Fig. 6 is a saturated pixel expansion area template;

Fig. 7 is the evaluation function curve of the defocus sequence.

detailed description

The present invention is an auto-focus evaluation function for an area containing saturated pixels. The function adopts a method of using a saturated pixel template to act on the gradient value matrix, and removes the influence of saturated pixels on the auto-focus evaluation function, and has good unbiasedness and single peak. performance, sensitivity, and stability.

Describe in detail below in conjunction with accompanying drawing and example:

Fig. 1 is a simplified flowchart of the method of the present invention. In this embodiment, a group of real-shot images with a number of steps of 42 is taken as an example. FIG. 2 is a schematic diagram of a defocus sequence of 5 real-shot images randomly selected from among the 42 pieces and the focus window area selected in this embodiment, where ( a)-(e) are the pictures obtained by the motor at the 1st step, 11th step, 21st step, 31st step, and 41st step, respectively, and the third position is the clear image of the quasi-focus position. Figure 3 is The actual scene diagram at the in-focus position (the third position in this example), and Figure 4 is the focus window diagram of the in-focus position. Figure 7 shows the acquisition process of this group of real shots. The motor is driven from the nearest end to the farthest end with a fixed step, so that the evaluation function curve shown in Figure 7 is drawn by calculating the evaluation function value of the focus window. The evaluation function curve calculation steps of the inventive method are as follows:

(1) Select a 200×200 focus window _Ic containing a light source in the auto-focus image I shown in Figure 3 containing a light source, as shown in Figure 4, use the gradient absolute value operator to calculate the focus window _Ic The gradient value at each pixel forms the gradient value matrix G of the focus window;

(2) Normalize the gray value of _Ic , and perform binarization through the threshold T to obtain the binary image M, and extract the saturated pixel area template _BIc :

BIc＝(Ic>T) (1)

Where BI _c reflects the area where the saturated pixel is located, as shown in Figure 5, where the white area is the saturated pixel area;

(3) Expand the template BI _c of the saturated pixel area by means of morphological expansion to obtain the expanded area template BI _c ' of the saturated pixel, and the expanded binary image is M';

Among them, Rid is a structural element of R×R; is an expansion operation, and M is a binarized saturated pixel expansion region template, as shown in Figure 6;

(4) Invert the expanded binary image M' to obtain the inverted binary image and calculate the binary map , the number N of pixels with a pixel value of 1;

(5) Binary image Act on the gradient value matrix G of the focus window to obtain the gradient information matrix W at the unsaturated pixels:

W is the gradient information matrix at unsaturated pixels, that is, the influence of saturated pixels is removed;

(6) Calculate the sum S _w of all gradient element values of the gradient information matrix W at the unsaturated pixel, and obtain the final definition evaluation function index Metric for desaturated pixels:

Metic＝Sw/N (5)

Among them, i and j are the rows and columns in the W template respectively, W(i, j) is the gradient value at the i-th row and j-th column in the gradient information matrix W, _a and _b are the length and length of the focus window I _c width.

Repeat the above steps to perform sharpness evaluation calculations on the 42 obtained defocus sequence images, and obtain the evaluation function curve of the focus window of the group of defocus sequences, as shown in Figure 7. It can be seen from this that the evaluation method of the present invention has good unbiasedness, unimodality, sensitivity and stability.

Claims (3)

1. A method for evaluating the quality of an autofocus image containing a light source scene, characterized in that the method may further comprise the steps: (1) Randomly select a focus window I _c containing a light source in the auto-focus image I containing a light source scene, and use the gradient absolute value operator to calculate the gradient value at each pixel in the focus window I _c to form the gradient value of the focus window matrix G; (2) Normalize the gray value of _Ic , and perform binarization through the threshold T to obtain the binary image M, and extract the saturated pixel area template _BIc : BI _c ＝1,I _c >T BI _c ＝0, I _c ≤T (3) Expand the template BI _c of the saturated pixel area by means of morphological expansion to obtain the expanded area template BI _c ' of the saturated pixel, and the expanded binary image is M'; BI _c '=BI _c ⊕ Rid (2); Among them, Rid is the structural element of R×R; ⊕ is the expansion operation; (4) Invert the expanded binary image M' to obtain the inverted binary image and calculate the binary map , the number N of pixels with a pixel value of 1; (5) Binary image Act on the gradient value matrix G of the focus window to obtain the gradient information matrix W at the unsaturated pixels: <mrow><mi>W</mi><mo>=</mo><mi>G</mi><mover><mi>M</mi><mo>~</mo></mover><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>3</mn><mo>)</mo></mrow><mo>;</mo></mrow> (6) Calculate the sum S _w of all gradient element values of the gradient information matrix W at the unsaturated pixel, and obtain the final definition evaluation function index Metric for desaturated pixels: <mrow><msub><mi>S</mi><mi>w</mi></msub><mo>=</mo><munderover><mo>&amp;Sigma;</mo><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mi>a</mi></munderover><munderover><mo>&amp;Sigma;</mo><mrow><mi>j</mi><mo>=</mo><mn>1</mn></mrow><mi>b</mi></munderover><mi>W</mi><mrow><mo>(</mo><mi>i</mi><mo>,</mo><mi>j</mi><mo>)</mo></mrow><mo>-</mo><mo>-</mo><mo>-</mo><mrow><mo>(</mo><mn>4</mn><mo>)</mo></mrow><mo>;</mo></mrow> Metric = S _w /N (5); Among them, i and j are the rows and columns in W respectively, W(i, j) is the gradient value at row i and column j in the gradient information matrix W, and a and b are the length and width of the focus window _Ic respectively . 2. The method for evaluating the quality of an auto-focus image of a scene containing a light source according to claim 1, wherein the threshold T in the step (2) ranges from 0.6 to 0.8. 3. The method for evaluating the quality of an autofocus image containing a light source scene according to claim 1, wherein R in the step (3) is 3.