Digital Communication
5.9 Digitizing Imagery
An important issue in net-centric warfare is the transportation of imagery from the point of origin to the point at which an operator or other decision-maker needs to access the information carried in the imagery. The imagery can be from a large part of the electromagnetic spectrum: visible light, infrared (IR), or ultraviolet (UV).
There are two basic approaches to the capture of imagery. One way is to scan an area using a raster scan as shown in Figure 5.31. In this technique, a single sensor (IR, UV, or visible light) (or set of sensors) is directed through the angular area of interest. The spacing of the lines in the raster is close enough to provide the required resolution of the picture in the vertical dimension. The horizontal resolution is determined by the angular movement between the samples of the data from the sensor. In analog video, this sampled data has a frame synchronization pulse at the beginning of each picture captured and a line synchronization pulse at the beginning of each line in the raster pattern. For commercial television (in the United States), there are 575 lines in the raster and 575 samples taken per line. Every second line (alternating) is sent 60 times per second. This captures 30 full pictures per second. In Europe, there are 625 raster lines and 625 samples per line. Every second line is sent 50 times per second, yielding 25 full pictures per second. In either case, this allows full-motion video because the human eye can only see a new picture 24 times per second. This analog video signal requires a bandwidth of just under 4 MHz in full color. By digitizing the output of the scanned sensor, a digital video signal is produced.
Figure 5.32 shows the other approach to capturing imagery data. In this case, there are a number of imagery sensors in an array. Each sensor captures one pixel of the picture. The outputs of these sensors are sequentially sampled and digitized to form a serial digital signal suitable for transmission. The bit rate of the digital signal is determined by the formula: Bit Rate Frames per Second × Pixels per Frame × Bits per Pixel A standard, full resolution digitized video signal has 720 by 486 pixels per picture with 16 bits for each pixel. This makes 720 × 486 × 16 bits per picture. In the United States, with 30 frames per second, this requires a bit rate of 167,961,600 bits/sec.
Figure 5.31 If imagery is sensed using a raster scan, the intensity of each color in each pixel is digitized into a serial bit stream.
Figure 5.32 If the imagery sensor has a sensor array, the intensity of each color is digitized for each pixel and output
as a serial bit stream.
In Europe, with 25 frames per second, the required bit rate is 139,968,000 bits/sec. The type of modulation carrying this digital data could require a great deal of link bandwidth. We will discuss various ways of reducing this data rate.
5.9.1 Video Compression
There are various basic measures that will reduce the required bandwidth. One way is to transmit analog video. Unfortunately, this option has the disadvantages that analog signals are very difficult to securely encrypt and their quality can be severely reduced if transported over long distances requiring multiple transmissions. If digital video is used, the required data rate (hence, bandwidth) can be reduced using several techniques:
• Reduce the frame rate.
• Reduce the data density (i.e., reduce the resolution).
• Reduce the angular area of coverage (with the same resolution).
• Take advantage of the fact that the eye sees luminance (brightness) at twice the resolution of chrominance (color). This allows full color with 8-bit resolution per color to be captured with only 16 bits per pixel.
• Use digital data compression software.
There are three basic digital compression techniques:
• Direct cosine transform compression (DCT) writes a digital word to describe an 8 × 8 section of the picture captured. This is a very mature technique. As the SNR of the received digital signal degrades, the picture breaks into square blocks. A single bit error will take out 64 pixels and under some circumstances can take out a whole picture, which can require multiple frames to resynchronize. Therefore, systems using DCT compression must usually incorporate forward error correction.
• Wavelet compression performs a series of highpass filter operations on the picture, replacing a series of 1s with a single 1. After repeating this operation 10 or 12 times, a compressed digital representation of the whole picture is generated. With this approach, each bit error has the effect of slightly blurring the whole picture. This means that, in general, forward error correction is not advantageous.
• Fractal compression is a process in which the picture is divided into geometric shapes and a digital bit stream is generated to describe the density, color, and placement of each shape. This technique requires a great deal of memory and processing power. The performance of this compression technique is comparable to that of DCT and wavelet compression but has the advantage of allowing significant enlargement.
Each of these techniques reduces the data rate that must be transmitted, thereby reducing the required link bandwidth. All three techniques compress each frame of video, which allows efficient editing and analysis to recover information from the digital data. The compression ratio depends on the required quality of the recovered video, but ratios of 30 to 50 are usually discussed. Temporal compression involves removing redundant data from frame to frame. It is possible to achieve very high compression ratios with this compression approach. The disadvantage is that digital editing becomes very difficult.
5.9.2 Forward Error Correction
By encoding transmitted digital signals with additional bits, it is possible to detect bit errors up to some limit and to correct those bit errors at the receiver. The more additional bits are incorporated, the more bit errors can be corrected. These additional bits increase the transmitted bit rate, hence the required link bandwidth.