Image Acquisition

From Streaming 411

1 Image acquisition

Image acquisition

Practically every process can be broken down into 3 basic steps - input, processing and output, so let's go through these with regards to digital video.

Cameras

The digital camera is actually a semi-digital device. First, it has to convert light (analog information) to electric signals (still analog) which are then digitized by means of sampling. Sampling is a key term used in any digital medium and it refers to the fact that the digital realm is finite - to deal in numbers, we have to set certain boundaries.

Sampling

“Sampling is the process of examining a value of a continuous function at regular intervals. We might measure an analog waveform every microsecond or the brightness of a black and white photograph every one-hundredth of an inch, horizontally and vertically. If we define the precision we need and then do the sampling job properly, we ca subsequently take those samples and reconstruct the original to that degree of defined precision.”

It’s just another word for measuring and a good example of a sampler is actually a theromometer. By measuring the temperature, we’re effectively sampling the temperature data at that given time. The weather doesn’t change very rapidly, so if we’re measuring the temperature outside, we could easily choose a sample rate of 1/hour or eve 1/day. What should be the sample size of a regular outdoor thermometer? For that we have to make a guess at what might be the maximum and minimum temperature we might ever measure, preferable with extra “padding” - possibly something like -50 to +50. But what if the temperature is 20,6 degrees? If we want to handle those values, we have to set the smallest increment. 1/10 (0.1 degrees) should probably be enough. That leaves the resolution of our thermometer at 10… and altogether we need 1000 descreet values, or 10 bits (210 = 1024) to describe the whole range of possible values

This is a little misleading because temperature is a signed number, something we don’t usually use in digital imaging.

The sample rate means how many times we measure a signal in each second. To get an optimal signal, we will need to use a sampling rate that's at least twice the maximum frequency of the source signal. The maximum frequency bandwidth of a TV signal is 6 Mhz so 13,5 Mhz is the standard frequency for standardizing PAL or NTSC video (although more and more hardware manufacturers use 27 Mhz). The ADC is practically a single chip, like the {National Semiconductor ADC08xx series}.

ADC converters work with one line at a time, let's find out how many samples will that leave for each line of video here in Europe? We can find it out with some easy algebra:

13500000 / 50 / 576 = 486

To use this sampled data, we must next put it into numbers - first, define the minimum and maximum values allowed, then divide the area into equal parts. This process is called quantization.