CN108282620B - Image color correction method comprising traffic signal lamp - Google Patents

Abstract

The invention discloses a method for correcting the color of an image containing a traffic signal lamp, which comprises the following steps: and calculating a truncation coefficient for each pixel in the target image data, and adjusting each channel data through the truncation coefficient so that each channel data does not exceed the dynamic range after the gain is amplified. According to the invention, the original image is acquired by reducing the exposure time and simulating the gain, the color component proportion of each pixel is recorded, and the color component proportion of the overexposed pixel is kept unchanged when the digital gain of the image is increased, so that the problems of yellow and white red light caused by over-brightness of the red light are solved without affecting the overall brightness of the image.

Description

Image color correction method comprising traffic signal lamp

Technical Field

The invention relates to the field of image processing, in particular to a method for correcting images shot with red traffic signal lamps.

Background

The electronic police is a modern traffic law enforcement system for automatically recording traffic violations such as red light running, reverse running, overspeed running and the like of a motor vehicle by using a high-definition camera, and GA/T496-2009 general technical conditions of an automatic red light running recording system puts requirements on recording pictures of the electronic police system: the picture definition can meet the requirement of manual identification of the number plate of the vehicle, and the manual judgment of the red light signal cannot be influenced by color distortion such as whitening and halo of the red light signal. "

Traffic signal lamps can be generally classified into incandescent lamps, halogen lamps and Light Emitting Diodes (LEDs) according to the difference of light sources, and the LED traffic signal lamps are widely applied in recent years due to the advantages of low power consumption, high brightness, no flicker of direct current and the like. The industrial camera is used as a core component of an electronic police recording system, and due to the limitations of signal lamp brightness, external light, shooting angle, self dynamic range and the like, in the night shooting process, under the premise of not influencing the brightness of surrounding scenes, the red signal lamp is easy to yellow, whiten and even completely overflow, and because the brightness of the LED signal lamp does not change along with time, the brightness of the signal lamp at the moment of shooting cannot be changed by adopting a camera and a synchronous mode thereof.

At present, most of the conventional red light correction methods adopt an image processing algorithm for reducing the exposure time of a camera, changing the brightness of ambient light, or replacing a red light region. The method in which the ambient light brightness is changed by reducing the exposure time of the camera or by adding an optical structure suppresses the discoloration of the red light, but causes the ambient brightness to decrease at the same time, making it difficult to meet the requirements. However, the method of identifying the red light region and then coloring the region makes the red light region appear hard and lack of reality, as shown in fig. 1, the picture obtained by the processing method is a picture, and it can be seen from the picture that the signal lamp has been processed to be deformed, and such an image is processed in a later stage, so that the image is relatively natural and has obvious fake marks, and is not allowed to be used for punishing evidences in principle.

Disclosure of Invention

The applicant has made intensive studies with respect to the problems of the conventional signal lamp correction method. The applicant has found that the prior art methods of colour correction for traffic lights do not address the cause of the colour change of the signal, and indeed none of the prior art solutions has found a root cause of the colour change of the red signal. The processing mode in the prior art is based on the processing of the whole image, but the invention provides a correction mode based on the dynamic range of pixels in the image, and the problem of yellowing of a red signal lamp is solved microscopically.

The invention provides a color overflow unbalance correction method based on original image restoration, which increases the digital-to-analog conversion precision and furthest ensures the original color proportion information of an image by reducing the exposure time of an acquisition system. When the system digital gain is adopted to increase the overall brightness of the image, the color proportion of the red overexposure pixel points is ensured, and therefore the problem that the red signal lamp is over-bright to cause partial yellow and white is solved. Specifically, the invention provides an image color correction method comprising a traffic signal lamp, which is characterized by comprising the following steps: and calculating a truncation coefficient for each pixel in the target image data, and adjusting the data of each channel of each pixel through the truncation coefficient so that the data of each channel does not exceed the dynamic range after the gain is amplified.

In a preferred implementation, the method comprises the steps of:

step S1: acquiring an image with a traffic signal lamp shot by a sensor;

step S2: performing analog-to-digital conversion on the acquired image to obtain corresponding image data;

step S3: respectively acquiring R, G, B channel pixel values of each pixel in the image data;

step S4: judging whether the R channel pixel value has amplification overflow or not;

step S5: calculating a truncation coefficient in different cases based on the overflow condition judgment in step S4;

step S6: and (4) multiplying the R, G, B channel pixel value of each pixel point by different truncation coefficients for differential amplification.

In another preferred implementation, the method further comprises shortening an exposure time of a sensor used to photograph the traffic signal light.

In another preferred implementation, the method further comprises reducing an analog gain of the analog-to-digital conversion module.

In another preferred implementation, the method further performs the amplification overflow determination by:

if Pr>2^N1, determining that amplification overflow exists; if Pr<＝2^N1, then it is determined that there is no amplification overflow, where: "Pr" represents the gray value of the red channel component of the pixel; "N" denotes D/A conversion accuracy(ii) a "Gain-number" indicates the Gain amplification factor.

In another preferred implementation, the method further performs truncation coefficient calculation by:

for the case of amplification overflow, P _ Gain is 2^N/Pr；

For the case where there is no amplification overflow: p _ Gain < ═ Gain, where P _ Gain represents the digital Gain value of the current pixel.

In another preferred implementation, the traffic signal lamps include at least a red signal lamp.

In another preferred implementation, the method is performed by an FPGA.

Technical effects

The invention collects the original image by reducing the exposure time and the analog gain, records the color component proportion of each pixel, keeps the color component proportion of the overexposed pixel unchanged when the digital gain of the image is amplified, and solves the problems of yellow and white red light caused by over-brightness of the red light without influencing the overall brightness of the image.

Compared with the traditional red signal lamp correction method, the method starts from the source of color proportion unbalance caused by dynamic range attenuation, and solves the problem that the environmental brightness and the red light color cast cannot be combined together on the premise of not losing the overall brightness of the image by retaining the original image color proportion information, so that the red light information is more real, and the problem that a pure post-coloring picture cannot be used as law enforcement evidence is effectively solved.

Drawings

FIG. 1 is a diagram: in the prior art, a region replacement processing mode is adopted to process the processed image processing result;

FIG. 2 is a diagram of: a dynamic range chain of image acquisition and display processes;

FIG. 3 is a diagram of: the color channel ratio of the pixel;

FIG. 4 is a diagram of: color channel ratio after dynamic range decay;

FIG. 5 is a diagram: setting a red light correction link of the camera;

FIG. 6 is a diagram of: a corrected color channel ratio after dynamic range attenuation;

FIG. 7 is a diagram of: images that have not been corrected by the method of the invention;

FIG. 8 is a diagram of: the image is corrected by the color correction method of the invention.

Detailed Description

The invention is described in detail below with reference to the drawings and the embodiments thereof, but the scope of the invention is not limited thereto.

The principle of the traffic signal color correction method of the present invention and the specific process of the correction will be described in detail below.

1. Dynamic range attenuation

In the research and development process, the applicant notices that the scene is gradually attenuated in the dynamic range in the whole process from the collection to the display of the camera, and the dynamic range of the whole system is determined by a link with the lowest dynamic range, as shown in fig. 2. For an industrial camera, the dynamic range of each pixel value in the sensor is determined by the characteristics of the sensor, the dynamic range of the pixel after ADC conversion by the analog-to-digital conversion device depends on the sampling width (for example, 14-bit AD is adopted) and the analog gain, the dynamic range of the pixel of the image processing part depends on data interception caused by the digital gain, and the dynamic range of the pixel of the display part depends on the response characteristics of the display.

2. Red light color cast due to pixel dynamic range attenuation

As shown in fig. 3, the color channel value ratios of the three pixels P1, P2, and P3 in the figure are not attenuated by any dynamic range, and if the three pixels are attenuated by a series of dynamic ranges and then truncated from the positions of the dotted lines, the color channel value ratios of the three pixels are as shown in fig. 4, it can be seen that the two pixels P1 and P2 have serious color channel ratio imbalance, which indicates that the color change occurs in the image.

The reason for this is that the red signal light is: due to the attenuation of the dynamic range of pixels in the industrial camera, overflow cut-off occurs in a red channel in the attenuation process, so that the proportion of color channels is unbalanced, and the phenomenon of color cast of red light occurs.

3. Adjustment of exposure time, analog gain, and digital gain

In the dynamic range chain of fig. 2, in order to prevent the dynamic range attenuation after AD and further cause the color channel scale imbalance, so that each channel of the stage can fall below the dotted line of fig. 3, the applicant proposes a new systematic correction method from image acquisition and analog-to-digital conversion to gain amplification.

Specifically, applicants first adjust the exposure time for the image sensor to capture the image, i.e., shorten the exposure time of the image sensor to reduce the dynamic range attenuation of the pixels in the image sensor (sensor); for example, the exposure time is adjusted to 1ms or 500us

Then, when performing analog-to-digital conversion, the applicant adjusts the analog gain value of the analog-to-digital conversion module, that is, the analog gain value of the analog-to-digital conversion is reduced to reduce the attenuation of the dynamic range of the pixels in a digital-to-analog conversion (AD) link; for example, the analog gain value is set to 0DB (amplification factor of 1)

The digital image data output by the AD enters the red light correction module, and because the exposure time and the analog gain value of the sensor and the analog-to-digital conversion module are reduced, in order to ensure the image sharpness, the applicant adjusts the digital gain to compensate, that is, increases the digital gain to ensure the overall brightness of the image, but this will cause the dynamic range of the pixel of the stage of digital amplification to have a larger attenuation, so the applicant designs a new overexposure pixel processing mode at this stage to perform overexposure processing, as shown in fig. 5.

4. Correction process

(1) It is determined for each pixel whether there is magnification overflow of the red channel, e.g., whether the magnified pixel gray scale exceeds 4095 for a 12-bit digital image

Overflow condition: pr (Pr) of>2^N/Gain-1

No overflow condition: pr (Pr) of<＝2^N/Gain-1

Wherein: pr-pixel red channel component gray value; n: D/A conversion precision; gain-digital Gain amplification. (Note: Gain is magnification factor, N is digital-to-analog conversion accuracy of digital image, for example, 12-bit image, N is 12)

(2) Calculating truncation coefficients

Overflow condition: p _ Gain 2^N/Pr

No overflow condition: p _ Gain ═ Gain (normally set Gain)

Wherein: p _ Gain represents the digital Gain value of the current pixel

(3) Digital amplification process

Pr’＝Pr*Gain

Pg’＝Pg*Gain

Pb’＝Pb*Gain

Wherein: pr ', Pg ', Pb ' -processed RGB component gray scale value of the pixel

Taking a 12-bit original image as an example, the 12-bit image needs to be amplified by 4 times and then output by 8 bits (i.e. 2-9 bit image data is taken), the processing method is to first determine whether the 12 red channel value of each pixel is greater than 1023, if so, the truncation coefficient is calculated:

p _ Gain is 4096/Pr, where Pr is the red channel value of the pixel.

If less than or equal to 1023, a truncation coefficient is calculated:

P_Gain＝4；

and then, the green-blue channel numerical value of the same pixel point is multiplied by the truncation coefficient of the pixel point for distinguishing and amplifying, so that the overall brightness of the image is improved, the color channel proportion of the red light region is kept unchanged, and the original color is kept, as shown in fig. 6.

TS Intelligent industrial camera

Because the correction method of the invention has small calculation amount and good correction effect, in an ITS intelligent industrial camera, the correction method of the invention can be completed by using an FPGA, and images before and after correction are as shown in figures 7 and 8. The overall brightness of the images in both figures is comparable, but in fig. 8 the color information of the red light is well preserved. Although the color picture is converted into a gray image in consideration of the requirements for the drawings in the patent, it can be seen that the display effect of the red light region in fig. 8 is significantly better than that of fig. 7 even from the gray image.

The foregoing is considered as illustrative and not restrictive, and all changes that come within the spirit and scope of the invention are intended to be embraced therein.

While the principles of the invention have been described in detail in connection with the preferred embodiments thereof, it will be understood by those skilled in the art that the foregoing embodiments are merely illustrative of exemplary implementations of the invention and are not limiting of the scope of the invention. The details of the embodiments are not to be interpreted as limiting the scope of the invention, and any obvious changes, such as equivalent alterations, simple substitutions and the like, based on the technical solution of the invention, can be interpreted without departing from the spirit and scope of the invention.

Claims (2)

1. A method of color correcting an image containing a red traffic signal, the method comprising: calculating a truncation coefficient for each pixel in the target image data, and adjusting data of each channel of each pixel through the truncation coefficient so that the data of each channel does not exceed a dynamic range after gain after amplification, wherein the method comprises the following steps:

step S1: acquiring an image with a traffic light shot by a sensor, wherein the exposure time of the sensor for shooting the traffic light is shortened during shooting;

step S2: performing analog-to-digital conversion on the acquired image to acquire corresponding image data, and reducing analog gain of an analog-to-digital conversion module during analog-to-digital conversion;

step S3: respectively acquiring R, G, B channel pixel values of each pixel in the image data;

step S4: judging whether the R channel pixel value has amplification overflow or not;

step S5: calculating a truncation coefficient in different cases based on the overflow condition judgment in step S4;

step S6: multiplying the R, G, B channel pixel value of each pixel point by different truncation coefficients for distinguishing and amplifying;

the method also performs an amplification overflow determination by:

if Pr>2^N1, determining that amplification overflow exists; if Pr<＝2^N1, then it is determined that there is no amplification overflow, where: "Pr" represents the gray value of the red channel component of the pixel; "N" denotes digital to analog conversion accuracy; "Gain" represents the Gain magnification;

the method also performs truncation coefficient calculation by:

for the case of amplification overflow, P _ Gain is 2^N/Pr；

For the case where there is no amplification overflow: p _ Gain < ═ Gain, where P _ Gain represents the digital Gain value of the current pixel.

2. The image color correction method according to claim 1, characterized in that the method is performed by an FPGA.

Images