The impact of CCD camera quality on network cameras

The impact of CCD camera quality on network cameras
Abstract: The quality of the CCD camera used in network cameras (network camera servers, video servers, network video converters, etc.) has a great impact on the bandwidth. Using high-quality CCD cameras can greatly reduce the bandwidth of network cameras.

1. Problem Statement

The development of network camera technology, especially the popularization of broadband networks, has made network-based network monitoring applications popular. The advantages of flexible monitoring and control configuration, close integration with information networks, and long-distance remote monitoring are incomparable to traditional monitoring. How to correctly master and reasonably apply network cameras is a new issue facing users.

The main problems in the actual application of network cameras are, first, image compression technology problems; second, network bandwidth data transmission problems. The essence of solving the bandwidth problem is how to transmit higher quality images on limited bandwidth, and of course, there is also the problem of improving the management capabilities of computer operating system software (decompression); the development of image data compression technology and the application of high-speed central processing units are the basic technologies for solving the image compression problem of network cameras. The development of many technologies (especially MPEG4) has undoubtedly enabled our current network cameras to achieve a higher image compression ratio than before. But are these enough? The answer is no.

2. Improving the quality of the original reference image is the key to image compression

Under certain compression technology and certain application environment, what factors affect the bandwidth of network cameras? This is whether the CCD camera used in the front end of the system can send out stable and clean images, that is, whether each frame is used as a newly added original reference image or as an unchanged compressible variant image. This factor is often overlooked. After the unstable and unclean image is input into the network camera, the network camera recognizes it as a newly added original reference image and retransmits it. The originally unchanged compressible variant image is misjudged as a newly added original reference image, which is manifested as color deviation, color rolling, picture distortion or rolling, interference stripes, noise, etc. Practice has proved that these unstable and unclean images transmitted by network cameras should have been transmitted as images without variation, so the transmission bandwidth occupied by these image data must be very large. Why does this result occur? This needs to be understood from the principle of image data compression.

Image data compression is usually achieved through two main links: the first link is to compress the image size. The smaller the image size, the shorter the image data, thus achieving the purpose of compressing the data volume. The second link is to compress the image formation process, that is, in the process of continuous image transmission, we only transmit the changing part of the image, and leave the relatively static part untransmitted, so that the image data becomes shorter and compression is achieved, which also reduces the excessive bandwidth occupied.

For example, we monitor the situation in a room. When there are no people or animals moving around in the room, the image we see is still. At this time, the network camera does not need to repeatedly transmit so many identical images, so the bandwidth occupied by the still state is very small. When people or animals move around, the network camera only needs to transmit the changed part of the image, which greatly shortens the entire image data. Although the image data is shortened, it still faithfully records the entire situation in the room. The network camera mainly detects whether the grayscale (or color level) of the image changes when identifying whether the image is still or active. If there is no change, it is still, otherwise it is active. However, the factors that affect the grayscale change are not only moving people and objects, but also changes in light and the influence of the unstable and unclean image noise signals mentioned above. In the eyes of the network camera, even if it is a still room, it also considers these unstable and unclean images as mutated images and increases the image data, which we do not need.

3. Use high-quality CCD cameras to ensure that the variation of images is reduced

Using high-quality CCD cameras to improve signal-to-noise ratio and achieve high-speed tracking white balance is the key to ensuring that the variability of the picture is reduced, thereby reducing the bandwidth of the network camera.

没有噪声的画面就是干净的画面，表现的指标就是信噪比高，相反，信噪声比不好的 CCD 摄像机的画面中充满噪点，而噪点在画面中的位置是没有重复性的，每一祯画面都不一样，在网络摄象机看来每幅都是全屏变化的图像，压缩后数据最大，占用的带宽也最多。实验表明，这样的画面数据比干净画面的数据要大 30% 以上甚至更多，这就是我们很多用户的网络摄象机带宽增大的主要原因之一。

Another parameter is the speed of the CCD camera tracking white balance. This parameter actually reflects the computing speed of the DSP chip used in the CCD camera. The computing speed of the DSP of an ordinary CCD camera is 100,000 times/second, while the computing speed of the DSP in a high-quality CCD camera can reach 10 million times/second, which is a difference of 100 times. What are the benefits of a faster DSP? The most obvious problem is color drift, because the computing speed directly affects the speed of its white balance tracking. We can use the fluorescent light environment to illustrate this problem.

The white balance speed of a general CCD camera is usually between 120 milliseconds and 160 milliseconds, which means that it takes about 6 to 8 fields (50 fields/second for PAL system) to complete a white balance change. This speed is definitely not able to keep up with the changes of fluorescent light sources. Fluorescent light changes on and off 100 times in 1 second, which means it changes once every 20 milliseconds. Obviously, general cameras cannot keep up with this change speed. Therefore, in the environment of fluorescent lighting, the image will produce color cast and color drift, and the originally static picture will become a "moving" image with periodic color changes. In this way, the network camera will naturally be deceived. However, if the DSP of the CCD camera has a fast computing speed, such as balancing a white balance change in only 1 millisecond, a very stable image can be obtained. The lesson from this example is that as environmental protection and energy-saving requirements continue to increase, the application of pulsed lighting will become increasingly widespread. In such a situation, the network camera will have to meet the DSP computing speed requirements of the CCD camera it uses. In other words, the requirement for the DSP computing speed of the CCD camera will become an indicator of network monitoring (the same applies to hard disk recorders).

From the above analysis, we can see that high-quality CCD cameras (providing a signal-to-noise ratio of 52db-60db and a DSP computing speed of 10 million times/second) can greatly improve the application quality of network cameras.

Of course, another indicator that affects the image quality of CCD cameras is the CCD sensitivity index. CCD cameras with high sensitivity index can improve the signal-to-noise ratio when the illumination is low.

It is worth noting that the definition index of CCD cameras has little effect on the image quality, but has a great impact on the bandwidth utilization of network cameras. The higher the definition of the image, the more storage space it occupies. Blindly pursuing the high definition of network cameras not only increases the cost of use, but also increases the transmission, processing and storage data of the image, which is an unscientific choice.