Image processing in CVS systems.

Image processing in CVS systems.

In CVS computer television systems, the amplitude and spectral characteristics of images in each frame are analyzed. The results of the analysis are then used to improve the image, the operation of motion and activity detectors, deinterlacing, eliminating redundancy, compression and transmission of images over the network.

Digital video recorders (DVR), hardware compression systems, simple computer systems do not have additional image processing.

Image Enhancement
Improving the quality of images in CVS systems built on computers is one of the main areas of activity of the developers of the New Technologies company.
Thanks to digital processing of the video signal, the contrast and brightness are automatically adjusted to optimal values, backlighting in the image is eliminated (BackLight function), manual and automatic gamma correction is performed, which is especially important for low-contrast and unevenly illuminated images (twilight, fog, shadows, etc.).
For a more complete analysis of the image, it is necessary to enter it with the maximum resolution both horizontally and vertically.
For a clear understanding of this issue, we will provide a short explanation.
The image resolution is determined by the entire input path, i.e. the camera, cable, electrical path, and the number of digitized pixels in a line.
When digitizing 704 pixels, even the theoretical resolution cannot exceed 528 TVL, while when digitizing 896 pixels (which is only possible in CVS systems), the theoretical resolution is 672 TVL. Higher resolution can be achieved by improving the signal-to-noise ratio (the «image noise reduction» function in CVS systems) and software control of the ADC filters (Fig. 1a, 1b).

– Under normal illumination, the resolution in CVS systems is 27% higher than in other systems.
Under low illumination in a noisy image, the resolution in CVS systems can exceed the resolution of other systems several times.
There are no analogs of the noisy video signal restoration function in other systems.

Digitization 50 frames per second

Digitalization in conventional systems is 25 fps when connecting one camera to the ADC and 8–16 fps in multiplex mode (for several cameras connected to one ADC).
Mathematical image restoration allows digitizing frames with a frequency of up to 50 fps both from one camera and in multiplex mode (in CVS systems, the multiplexing frequency of asynchronous cameras is 40-50 fps per ADC). Moreover, for the moving part of the image, deinterlacing is performed in frames, which allows eliminating the «comb» on moving objects while simultaneously preserving the frame resolution for the still part of the image.
Of course, this seems paradoxical — digitizing 50 images per second with frame resolution, but this is true. This method, implemented, for example, for the «Akkord-16» model, allows you to receive 10 images per second with frame resolution and deinterlacing for each of the 16 cameras.
In multiplex mode, CVS systems operate 3-5 times faster than competing systems.
It is possible to digitize images with a frame resolution of up to 50 fps in both real and multiplex modes. There are no analogues of such a speed of digitizing images.

Motion detector, activity detector
Activity detectormonitors changes in the amplitude of the video signal in the entire unmasked field of view of the camera. The sensitivity of the activity detector is determined automatically and usually has a value of about 1-3% of the maximum amplitude of the video signal (the sensitivity threshold of the trained human eye to changes in amplitude on the monitor screen is 2-3%). The task of the activity detector is to:
exclude unchanging images from processing;
highlight images or objects in the image with changes that the human eye can detect;
provide information to the system to enable subsequent processing algorithms (digitization acceleration, filtering, deinterlacing, motion analysis, delta compression, recording, network transmission, etc.).

The motion detector has up to 16 independent rectangular motion detection zones (including intersecting ones) for each channel. The motion detector does not react to changes in illumination.
The task of the motion detector is:
determine the presence of movement of an object with given dimensions in the controlled area with a minimum of false alarms (elimination of noise, changes in illumination, etc.);
provide information to the system to enable subsequent event processing algorithms: enable accelerated recording of the event history or only the movement of the object; enable an alarm on the computer or additional posts, output additional information about the alarm; execute specified scenarios: track the target, enable the relay, etc.
The CVS detector has an extremely high sensitivity of 1–3%, which is the result of using automatic noise threshold correction and compensation for changes in illumination.
Using an activity detector and a motion detector allows for optimal use of processor resources, as well as saving disk space for archiving useful video information.
There are no analogs of the CVS activity detector and motion detector with such sensitivity.

Elimination of information redundancy
CVS systems implement several methods for eliminating redundancy of information based on the results of activity and motion analysis: delta compression, dynamic control of the sampling frequency in matrix switches, dynamic resolution control, control of dome cameras based on the results of motion analysis in the field of view of overview cameras.

Delta Compression
When guarding objects, stationary cameras are mainly used, and changes in the frame are local in nature and take up little space in the frame area.
Therefore, in a sequence of frames, only a small active part of the image is compressed and recorded, and rarely reference (background) frames (Fig. 2). This leads to a proportional reduction (tens of times) in the volume of the archive and data flows through the information transmission channels.
As a result, the information compression method developed and implemented in CVS systems works much more efficiently than other methods.
The omission of changed images is excluded, since the system's sensitivity threshold (1–3%) is higher than the eye's sensitivity and is adjusted automatically during operation.

Fig. 2. CVS delta compression Delta compression is fundamentally different from all known compression methods, such as MPEG-4 and H.264. Intermediate frames in delta compression are of the same quality as reference frames, not predicted and, as a rule, blurred. Moreover, MPEG*, H.*** algorithms are oriented to recording films, i.e. video fragments with small interframe changes over the area of ​​the entire frame. They are practically not applicable for multiplex systems.

Dynamic control of the sampling frequency in matrix switches
The matrix switch is connected to several ADCs, for example, to 4, as in the CVS_EMS 24 x 8E model (Fig. 3). Activity and motion are analyzed by one – the main channel. When activity or motion is detected, the cameras are automatically transferred for digitalization by additional ADC channels at the maximum frequency: for 3 cameras – 50 fps, for 6 – 17 fps, for 12 – 12.5 fps, but not less than 7 fps for all 24 cameras. From practical experience it is known that at a protected facility the average activity does not exceed 25%, and motion is even less. As a result, when the CVS system operates with a matrix switch “by activity” or “by motion”, the digitalization frequency reaches 17–50 fps for each camera.

Fig. 3 CVS_EMS 24 x 8E With a real performance of no more than 200 fps, the CVS_EMS 24 x 8E system is equivalent to a system with a digitization rate of up to 1200 fps. In addition:
an almost 6-fold additional compression of information is achieved,
the requirements for processor performance are reduced.
There are no analogues of CVS systems with dynamic control of the digitization frequency in matrix switches on the market.

Dynamic image resolution control
For monitoring images in multi-camera mode, a low resolution (e.g. CIF) is sufficient. When monitoring an image expanded to full screen, a high resolution (4CIF) is required.
In CVS, the maximum resolution is automatically enabled when opening an image to full screen. This feature is available on both the server and the client, for both analog and IP cameras.

Control of dome cameras based on images from overview cameras
Control of high-speed dome cameras based on target designations from overview cameras is implemented in the software module «CVS Virtuoz». Let us explain the principle of its operation.
The system must include one or more overview cameras and a high-speed dome camera. The coordinates of the object, as well as the distance to it, are determined based on images from the overview camera, then a command is given to the dome camera to show this object with the required zoom. The module has two modes: manual (MS) and/or automatic object tracking (AC). Since the coordinates of objects are determined based on images from overview cameras, the system can track several objects simultaneously.
How is information compression achieved in CVS Virtuoz? Obviously, when operating one overview camera and a dome camera with 10x zoom in the automatic detection and tracking mode, such a system can replace a system with 100 conventional overview cameras. Accordingly, the redundant information will be reduced by the same amount.

By the way …
All of the above image processing methods, with the exception of dynamic control of the digitization frequency in matrix switches, are applicable to both analog and network IP cameras. In addition, only in CVS do users have additional options for correct image orientation: depending on the camera position, the image can be rotated by 90, 180 or 270 degrees.