Bit Depth

Bit Depth . It is the amount of binary information that makes up an image, or range of colors and tones that the digital image will have, either in color or in black and white.

Summary

[ hide ]

  • 1 Bit depth samples
  • 2 Resolution Frequency
  • 3 Indexed Color
  • 4 High Resolution Color or HiColor
  • 5 True Color or True Color
  • 6 Source

Bit depth samples

What the bit depth tells us is: what size will each of the samples be, the larger the sample size, the greater amount of information we can have about each one, and logically our files will also be larger. The different depths that we find in audio are: 8, 16, 20, 24, 32 and 64 bits.

If we have 24 bits for each sample, what will happen when we reach bit number 25? When we exceed the number of spaces assigned to define our sample (wordlength) it means that there is no longer a way to interpret data to generate sound and what we obtain is what is known as a ‘noise bed’ to this phenomenon in which noise is created by ‘indefiniteness’, because we no longer have a way to express something beyond the allocated number of bits and it is called ‘sample truncation’. In practical terms this is the ‘Hiss’ and it is equivalent to the noise that we find in an analog tape, the difference is that in an analog environment the noise is produced by the friction between the tape and the playhead while in the digital environment the Noise is a product of the aforementioned sample truncation or wordlength truncation. And the more bits of wordlength we have, the lower the noise bed will be in volume. From this it also follows that the more bits there are in the wordlength, the additional information that we can capture will correspond to the fainter part of the sound. This phenomenon is known as ‘signal to noise range’ or signal to noise ratio. We express the volume in decibels (dB) and contrary to what we might think, this measurement does not correspond to volume but to pressure (force on a surface), In the particular case of sound, we refer to the air pressure above our ears and it is abbreviated dBSPL (in English deciBels Sound Pressure Level) and its values ​​go from 0 to (infinity). In turn, audio devices have their own way of ‘measuring’ the volume and they do it from 0 as a maximum value towards – or where the noise bed of the system begins and always in negative numbers, this represents the voltage necessary to produce a certain volume or rather a certain power; these values ​​correspond to full scale decibels or dBFS (decibels full scale). Within our topic, dBFS is what interests us, as this way we can calculate the noise bed for each bit depth, as well as some approximations for analog environments. In analog systems the noise bed is variable and in the digital environment the noise bed is exact. these values ​​correspond to full scale decibels or dBFS (decibels full scale). Within our topic, dBFS is what interests us, as this way we can calculate the noise bed for each bit depth, as well as some approximations for analog environments. In analog systems the noise bed is variable and in the digital environment the noise bed is exact. these values ​​correspond to full scale decibels or dBFS (decibels full scale). Within our topic, dBFS is what interests us, as this way we can calculate the noise bed for each bit depth, as well as some approximations for analog environments. In analog systems the noise bed is variable and in the digital environment the noise bed is exact.

Resolution Frequency

This is the other crucial component of the digital domain, but what does 44.1, 48, 96, and 192 KHz mean? In simple terms this means how many samples can be analyzed per second, if we have a depth of 24 bits at 48 KHz (which is the resolution of a DAT) it means that we can process 48000 samples per second of 24 bits each, the higher it is. the sampling frequency, the greater number of samples we will have and therefore we will be able to analyze in greater detail.

Until now, as in the case of bit depth, the higher resolution we have, the better, but in our case it is not the case and this is where we are going to solve the myth of the very high sampling frequencies that do not give us a benefit and yet the higher resolution higher cost to us.

The first reason is the limitation of the highest frequency to analyze:

How do we know that sound also has frequencies and of course this is also measured in Hertz (Hz). Frequency is defined in simple terms as the number of complete cycles per second that occur (a positive and a negative period), let’s see the following figure. The horizontal line in the middle represents time, in this case fractions of a second, the rectangles represent a sample with a resolution of let’s say 48 KHz, within each frame we find the positive phase of a wave and in the next one the negative phase of the same wave ,

Indexed color

For color depths less than or equal to 8, the pixel values ​​refer to RGB tones indexed in a table, commonly called the colorization creator box or palette. The tones in this table can be defined by convention or be configurable, depending on the application that defines it. Next, some color depths in the low range will be mentioned, as well as the amount of tones that they can represent in each pixel and the name that is given to the images or framebuffers that support them.

1 bit per pixel: 21 = 2 colors, also called monochrome or black and white. Compatible with IBM PC with MDA or HGC, early Macintosh, Atari ST in high resolution

2 bits per pixel: 22 = 4 colors, or CGA

3 bits per pixel: 23 = 8 colors – early home computer models like the ZX Spectrum and BBC Micro

4 bits per pixel: 24 = 16 colors, which is the minimum depth accepted by the EGA standard. Color Macintosh, Atari ST, Commodore 64, Amstrad CPC.

5 bits per pixel: 25 = 32 colors, as on the original Commodore Amiga chipset

6 bits per pixel: 26 = 64 colors, as on the original Commodore Amiga chipset

8 bits per pixel: 28 = 256 colors, also called VGA. Super VGA, Macintosh color, Atari TT, Commodore Amiga with AGA chipset, Atari Falcon030.

10 bits per pixel: 210 = 1024 colors, used in UHDTV.

12 bits per pixel: 212 = 4096 colors, some Silicon Graphics models, color NeXTstation, Commodore Amiga HAM mode.

High Resolution Color or HiColor

The 15- and 16-bit color depth values ​​are commonly called high resolution color or HiColor.

  • At the 15 bpp depth, 5 bits are used to encode the intensity of red, 5 for green and the other 5 for blue. With a depth of 15 bpp it is possible to represent 32 x 32 x 32 = 32768 colors in each pixel.
  • At the 16 bpp depth, 5 bits are used to encode the intensity of red, 6 for green and the other 5 for blue. The reason for this is that it is experimentally known that the human eye is more sensitive to the color green, and that it can discriminate more shades that vary slightly in green intensity. With the depth of 16 bpp it is possible to represent 32 x 64 x 32 = 65536 colors in each pixel. The size of the image will depend on the bit depth.

True Color or True Color

For the color depth of 24 bits per pixel, we speak of true color because the polychromy is close to what the human eye can find in the real world, since the human eye becomes unable to differentiate between a tone and another, if the difference remains within a certain minimum range. At 24-bit per pixel color depth, an entire octet is dedicated to representing the light intensity of each of the three primary shades of red, green, and blue, allowing each pixel to take 224 = 256x256x256 = 16,777. 216 different colors. When 32 bits are used to represent a color, a fourth channel called alpha is added to the scheme of 256 values ​​for each primary tone, which represents transparency. This value is used when two images must overlap.

Edwin Catmull proposed a series of operators for blending (also called blending) images. Because Catmull was developing Pixar’s animation system, he needed to automate image composition. Together with Marc Levoy (Levoy developed them on his own at Hanna Barbera Studios) they laid the foundations for color algebra. The most common operator is the OVER, which we present below: Given the colors A [r1, g1, b1, a1] and B [r2, g2, b2, a2], the result of mixing: A over B is:

A OVER B = [r1 * a1 + r2 (1 – a1), g1 * a1 + g2 (1 – a1), b1 * a1 + b2 (1 – a1)] Note that the order is important: A over B is different than B over A. This is why transparency should be applied in order. For example if we have three images A, B and C, where C is the background and A is the front image, the order is as follows: Final Image = A over (B over C) There are other important operators such as OUT, ADD , SUB and other more complex OVER formulations.

 

Leave a Comment

Scroll to Top