A Design Scheme for Low-light-level Video Processing System

Low-light-level video systems are increasingly used in night vision technology. However, due to the low target illumination, close target and background reflectance, and poor observation environment, the low-light-level video images obtained are always unsatisfactory. Under the current hardware production capacity in my country, low-light-level image intensifiers cannot achieve practical results. However, real-time image processing technology can quickly improve the image quality of low-light-level systems and increase the viewing distance of low-light-level video systems. Therefore, combining real-time image processing technology with image enhancement technology has become an important development direction to shorten the gap between domestic and foreign low-light-level video systems. This paper implements a real-time processing system for low-light-level video on the Da Vinci platform TMS320DM6446, which has important reference value for the design of low-light-level video systems in my country.

1 Low-light video enhancement technology

The most notable feature of low-light video is the obvious random flicker noise superimposed on the image. By analyzing the characteristics and noise of low-light video, the low-light video enhancement system can be designed from two aspects: increasing the contrast of low-light video and reducing the noise of low-light video. Due to the characteristics of the system, a simple and effective processing algorithm is needed. Histogram equalization is a classic algorithm for increasing image contrast, and the algorithm is simple and has good processing effect; Gaussian smoothing algorithm can effectively remove the high-frequency components of the image and retain the low-frequency components. These two algorithms can be used in combination with embedded processing platforms to achieve good results. This paper uses software design methods in the design to achieve image enhancement technology.

2 Low-light video system based on Da Vinci platform

The characteristic of the low-light video system is that it can realize the enhancement and denoising of the low-light video on the basis of real-time acquisition and real-time display to meet the requirements of real-time processing. Therefore, it requires strong real-time processing and data processing capabilities. Here, TI's Da Vinci platform (TMS320DM6446) with DSP+ARM dual-core processing functions is selected to meet the above requirements.

In the low-light video real-time processing system, the low-light video acquisition, processing and display process is through the CCD/CMOS camera or image decoder on the TMS320DM6446 platform. After real-time acquisition and digital conversion of low-light video, it is stored in the read buffer. Users can use the digital media processor of TMS320DM6446 to complete the real-time processing task of low-light video, and display the processed low-light video in real time through the display device on the embedded processing platform. The principle of the low-light video real-time processing system is shown in Figure 1.

Figure 1 Schematic diagram of low-light video real-time processing

In this low-light video real-time processing system, the low-light video real-time acquisition part is completed by video acquisition equipment and video decoding chip; the low-light video real-time processing part is completed by TI's TMS320 DM6446 dual-core processor; the low-light video real-time output system is completed by LCD display and video encoder.

[page]

2.1 Collection of low-light video images

The analog video digitization standard ITU-R BT. 656 in the paper is a new video digitization standard developed on the basis of ITU-R BT. 601 (CCIR-601). Since the CCIR656 video data stream uses 8 signal lines (the traditional method requires 19 signal lines) to transmit all image information and line synchronization, field synchronization, and even synchronization information, when the CCIR656 method is used for interface design, it is convenient for the real-time digital image processing hardware platform to be modularly designed with functional units as units.

Figure 2 Structure diagram of low-light video acquisition module

The analog low-light video is collected by CCD/CMOS video acquisition equipment. The video is in PAL or NTSC format. The analog video is converted into YCrCb format video through A/D conversion by decoding chip TVP5146 and BT.656 interface of CCD controller. At the same time, the collected low-light video is cached in the read buffer of DDRAM or SDR-AM through external memory interface (EMIF) so that the low-light video processing module can process the collected data.

2.2 Design of low-light video processing function

According to the video format, the component is then converted from RGB format to YUV format, and then processed in the embedded processing platform. After processing, the component is used to replace the Y component, and the grayscale image is converted into a color image for output and other operations. The structure of the video processing module is shown in Figure 3.

Figure 3 Structure diagram of low-light video processing module

[page]

After the acquisition module acquires a frame of low-light video, the video processing module obtains a frame of low-light video from the read buffer in DDRAM/SDRAM, enhances and denoises the low-light video in turn, and caches it again in the write buffer in DDRAM or SDRAM after processing for use by the display module.

2.3 Design of low-light video display function

In this system, we use Framebuffer technology and DaVinci's video processing subsystem VPSS technology to display images. Framebuffer is a driver interface in the Linux kernel that abstracts the display device into a frame buffer. Users can view it as an image of the display memory without having to worry about the location of the physical video memory, the paging mechanism and other specific details. Because these details are all completed by the Framebuffer device driver. The programmer only needs to map the image to be displayed to the process address space, and it will be displayed on the screen. The structure diagram of the display module is shown in Figure 4.

Figure 4 Structure diagram of low-light video display module

The video collected by the acquisition module is enhanced and denoised by the processing module, and then displayed using the FrameBuffer technology in a memory mapping manner.

3. Software Design of Low-light Video Real-time Processing System

Since DM6446 integrates ARM and DSP dual cores, the ARM side is the main control device, and the DSP side is used to process image processing, which greatly improves the processing capacity of the system. In the software design, this system can be mainly divided into three parts: low-light video acquisition module, low-light video processing module and low-light video display module. The system starts with the video acquisition device to collect low-light video, and digitizes the collected analog video image through the BT. 656 standard, and stores the digitized low-light video in the read buffer of SDRAM/DDBAM for use by the low-light video processing module; the low-light video processing module reads the low-light video from the read buffer, pre-processes the low-light video, extracts the fitting component that needs to be processed, and then enhances and denoises the component. After processing, the grayscale image is restored to a color image and written to the write buffer in SDRAMIDDRAM for display by the display module; the functional block interaction of the low-light video real-time processing system is shown in Figure 5.

Figure 5 Interaction diagram of functional modules of low-light video real-time processing system

[page]

4 System Testing

The implementation part was tested on the Da Vinci platform TMS320DM6446. On the basis of achieving real-time acquisition and real-time display, real-time processing was performed. The test results of each module of the system are as follows. The original low-light video collected is shown in Figure 6.

Figure 6 Original low-light video

Figure 6 is the original image after the low-light video system. It can be seen that the image contrast is relatively low and there is granular noise on the surface. The real-time test results of video acquisition based on the TMS320DM6446 platform are shown in Table 1. It can be seen that before adding the image processing algorithm, the video acquisition reached 25fps, meeting the real-time requirements. The CPU utilization in ARM and DSP and the video bit rate are shown in Table 1.

Table 1 Test results before video processing

Figure 7 is a real-time processing test of low-light video based on real-time acquisition. The image is processed by the enhancement algorithm in this paper on a frame of low-light video.

Figure 7 Enhanced image

Figure 7 shows the low-light video enhancement effect. The real-time test of low-light video processing is shown in Table 2. After acquisition and processing, the video is displayed at a speed of 24fpe, which basically meets the real-time requirements. At this time, the CPU utilization and video bit rate are shown in Table 2.

Table 2 Test results after video enhancement

From Figures 6-7 and Tables 1-2, it can be found that the system has been enhanced in image processing capability and system CPU utilization compared with the original system, reaching the processing capability of the low-light system.

5 Conclusion

The Da Vinci platform integrates a general-purpose processor and a digital signal processor in one dual-core. Compared with the previous single-core processing platform, it has higher performance and lower power consumption, and can meet the requirements in both computing power and control functions. The low-light system design scheme given in this paper has important reference value for the design of related video processing systems.