CN102054266A - Method for suppressing night low-illumination image noise - Google Patents

Abstract

The invention discloses a noise suppression method for low-illumination images at night. The specific method is as follows: firstly, judge and find out the low-illuminance image enhanced by a large gain, and perform nonlinear mapping on the noise-sensitive area in the low-illuminance image. The mapped image is subjected to median filtering again, and the final output image is the image after noise suppression. The invention adopts the method of combining nonlinear mapping and median filter to suppress the noise of the image, thereby achieving the suppression of the noise of the low-illuminance image while maintaining the visibility of the original image, and improving the monitoring picture quality of the night image.

Description

A Noise Suppression Method for Low Illumination Image at Night

technical field

The invention belongs to the technical field of all-weather video monitoring, and relates to a noise removal method for low-illuminance images at night, in particular to a method combining nonlinear mapping and a median filtering method to suppress random noise in low-illuminance images.

Background technique

For situations such as intelligent traffic monitoring that require all-weather monitoring of the monitoring field of view, when entering the night, because the illumination is too low, in order to see the scene clearly, it is necessary to enhance the low brightness with a large gain, which leads to the large gain of the monitoring screen. The noise is too strong, which greatly affects the display effect of the picture.

Contents of the invention

The purpose of the present invention is to provide a noise suppression method for low-illumination images at night, which can suppress noise under the premise of maintaining the visibility of the scene in the low-illumination image at night when the image noise is too strong due to large gain enhancement.

The technical solution adopted in the present invention is a nighttime low-illumination image noise suppression method, which suppresses the noise-sensitive area of the image through nonlinear mapping, and then performs a median filter on the nonlinearly mapped image to convert the remaining The noise is suppressed again, specifically as follows:

Step 1. Determine whether the read-in image is a low-light image enhanced by a large gain

，计算读入图像的均值：Let the image read in be

, calculate and read the image mean of:

             （1）

(1)

分别为图像的高、宽像素个数，where, where, is the average brightness of the image,

are the height and width pixels of the image, respectively,

，如果

，则判断其为大增益增强后的低照度图像；Set the image brightness value range given by the device as

,if

, it is judged to be a low-light image enhanced by a large gain;

Step 2. Non-linear mapping of low-light images

In step 1, it is judged that the area with rich noise distribution in the low-light image after large-gain enhancement is suppressed by nonlinear mapping method, and the suppressed image is set as ,but

(2)

为设备给定图像亮度值范围的上限；

为调整量；

为方差；

为调整系数，取值为in,

Specify the upper limit of the image brightness value range for the device;

is the adjustment amount;

is the variance;

is the adjustment coefficient, the value is

                          （3）

(3)

的均值；in, for

the mean value of

进行中值滤波，Step 3. Image after nonlinear mapping of step 2

Perform a median filter,

，即为噪声抑制图像，Step 4. output result image

, which is the noise-suppressed image,

If it is a color image, perform the above steps on the three color channels respectively to obtain a noise-suppressed image.

The low-illuminance image noise suppression method of the present invention adopts the combination of nonlinear mapping and median filter, and effectively suppresses the noise under the premise of maintaining a large gain to enhance the visibility of the scene.

Description of drawings

Fig. 1 is the filtering template schematic diagram that uses in the inventive method;

Figure 2 is the read image with noise;

Fig. 3 is the result image after the non-linear mapping process of step 2 is carried out to Fig. 2 by the method of the present invention;

Fig. 4 is the final result image of Fig. 1 after steps 2 and 3 are processed by the method of the present invention.

Detailed ways

The present invention will be described in detail below in combination with specific embodiments.

The noise suppression method of the low-illuminance image of the present invention, the specific operation method is as follows:

Step 1. Determine whether the read-in image is a low-light image enhanced by a large gain

，图2显示的是一个带噪声图像的具体实例，Let the image read in be

, Figure 2 shows a specific example of a noisy image,

的均值：Calculations read in images

mean of:

             （1）

(1)

为图像的亮度均值，

分别为图像的高、宽像素个数。in,

is the average brightness of the image,

are the height and width pixels of the image, respectively.

，如果

，则判断其为大增益增强后的低照度图像。  The range of image brightness values given by the device is

,if

, then it is judged to be a low-illumination image enhanced with a large gain.

Step 2. Non-linear mapping of low-light images

，则In step 1, the area with rich noise distribution in the high-gain enhanced low-illumination image is suppressed by nonlinear mapping method, and the suppressed image is set as

,but

(2)

为设备给定图像亮度值范围的上限，对于8bit位图，

；

为一个小的调整量，建议取

最大为

的1%；为方差，方差越大，则抑制噪声效果越好，但同时图像的细节破坏也越严重，因此，建议

；

为调整系数，用以保持非线性映射前后图像的亮度不变，in,

Specify the upper limit of the image brightness value range for the device, for 8bit bitmap,

;

For a small adjustment amount, it is recommended to take

up to

1%; is the variance, the larger the variance, the better the noise suppression effect, but at the same time, the more serious the detail damage of the image is, therefore, it is recommended

;

is an adjustment coefficient to keep the brightness of the image before and after nonlinear mapping unchanged,

                          （3）

(3)

in, for The mean value of can be calculated according to formula (1), and the effect of the image of the specific example in Figure 2 after this step is shown in Figure 3.

进行中值滤波，进一步抑制噪声，中值滤波步骤如下。Step 3. Image after nonlinear mapping of step 2

Perform median filtering to further suppress noise. The steps of median filtering are as follows.

1) Set the template of the median filter

，十字的四段直线端点距离中心点的长度为

，

为大于0的整数，由此使模板覆盖的像素个数为奇数个，如图1所示。

越大，抑制噪声效果越好，但同时景物细节的损失也越大，因此，建议取

。Using a cross-shaped template shape, the template size is

, the length of the four straight line endpoints of the cross from the center point is

,

is an integer greater than 0, so that the number of pixels covered by the template is an odd number, as shown in FIG. 1 .

The larger the value, the better the effect of suppressing noise, but at the same time the loss of scene details is also greater. Therefore, it is recommended to select

.

The role of the template is to determine the neighborhood range participating in the median filter, and not to process the edge part of the image that the template cannot fully cover.

的待处理像素放置在模板的中心位置，即取到的模板覆盖的邻域为

；2 pairs

of pending pixels Placed at the center of the template, that is, the neighborhood covered by the obtained template is

;

3) Sort all the pixels in the neighborhood covered by the template in ascending order;

的中值，即排在

+1位置上的像素值作为中值滤波后的值，设中值滤波结果图像为

，则4) Take the sorted neighborhood

the median value of the

The pixel value at the +1 position is taken as the value after the median filter, and the result image of the median filter is set as

,but

(4)

表示取中值操作。function

Indicates the median value operation.

，即为噪声抑制图像，对图2具体实例的图像经本发明处理后的效果如图4所示。Step 4. output result image

, which is the noise-suppressed image, and the effect of the image of the specific example in FIG. 2 after being processed by the present invention is shown in FIG. 4 .

If it is a color image, perform the above steps on the three color channels respectively to obtain a noise-suppressed image.

Claims (4)

1. A noise suppression method for low-illumination images at night, through nonlinear mapping, the noise-sensitive area of the image is suppressed, and then a median filter is performed on the image after the nonlinear mapping, and the remaining noise is suppressed again, It is characterized in that the specific operation is as follows: Step 1. Determine whether the read-in image is a low-light image enhanced by a large gain Let the image read in be

, calculate and read the image

mean of: (1) where, where,

is the average brightness of the image, are the height and width pixels of the image, respectively, Set the image brightness value range given by the device as

,if

, it is judged to be a low-light image enhanced by a large gain; Step 2. Non-linear mapping of low-light images In step 1, it is judged that the area with rich noise distribution in the low-light image after large-gain enhancement is suppressed by nonlinear mapping method, and the suppressed image is set as

,but

(2) in,

Specify the upper limit of the image brightness value range for the device;

is the adjustment amount;

is the variance;

is the adjustment coefficient, the value is

(3) in,

for

the mean value of Step 3. Image after nonlinear mapping of step 2

Perform a median filter, Step 4. output result image

, which is the noise-suppressed image If it is a color image, perform the above steps on the three color channels respectively to obtain a noise-suppressed image. 2. according to the suppressing method of nighttime low illumination image noise of claim 1, it is characterized in that, in described step 2: for 8bit bitmap,

The value is ;

The value is ≤ 1%;

. 3. according to the suppressing method of nighttime low illumination image noise of claim 1, it is characterized in that, the median filter step in the described step 3 is as follows: 1) Set the template of the median filter Using a cross-shaped template shape, the template size is ,

is the length from the end points of the four straight lines of the cross to the center point,

The value is an integer greater than 0, and the edge part of the image that cannot be completely covered by the template is not processed. 2) Will

of pending pixels

Placed at the center of the template, that is, the neighborhood covered by the obtained template is

; 3) Sort all the pixels in the neighborhood covered by the template in ascending order; 4) Take the sorted neighborhood

the median value of the The pixel value at the +1 position is taken as the value after the median filter, and the result image of the median filter is set as

,but (4) function

Indicates the median value operation. 4. The nighttime low-illumination image noise suppression method according to claim 3, characterized in that, in step 1) of step 3,

The value is 3.

Images