Bit Depth. Introduction

Transcription

1 Colourgen Limited Tel: +44 (0) The AmBer Centre Sales: +44 (0) Oldfield Road, Maidenhead Support: +44 (0) Berkshire, SL6 1TH Accounts: +44 (0) United Kingdom Fax: +44 (0) Bit Depth Introduction Before explaining about bit depth, an understanding of bitmap images is required and this is explained in detail, in the document entitled, Difference between Bitmap and Vector but below is a small section from said document. The complete document can be found using the following URL. Bitmap (AKA Raster) images are made up of pixels and normally captured from a digital camera or scanner. Pixels are small squares with a numerical value which depicts the colour of each pixel. These pixels are mapped into a grid and the different colour shades of the pixels together form the overall image. If the right ratio of size of image and the number of pixels which it contains are set correctly, then the image should be perceived as continuous tone, meaning that the square pixels cannot be seen with the naked eye. Bit depth is also commonly known as Colour depth or Pixel depth and as explained above each pixel has a value which depicts the colour and this is where bit depth comes into play. Even if you have a basic understanding of bits, bytes, Kbytes or megabytes (in terms of bitmap images we are only looking at bits and bytes, in which 8 bits make 1 byte), many people don t understand how it relates to the colour information stored within a pixel. It can quite literally mean the difference between a black and white image with no shades, to an image with millions of available colours and continuous tone perceived shading. Bits A bit (which gets its name from Binary Digit) is the smallest unit in terms of bit depth. Each bit is an electronic pulse which can either be on (represented by a 1) or off (represented by a 0). Binary data (also referred to as the Base 2 numbering system) uses the mathematical formula, two to the power of n, or 2 n. Bit depth refers to the number of colours which can be stored within each pixel and the most commonly used bit depths are 1-bit, 4-bit, 8-bit and 24-bit and by using the 2 n formula the number of colours for each level of bit depth can be calculated. This document will cover the differences between 1-bit, 8-bit, 24-bit, 32-bit and 48-bit 1-bit (Monochrome) 1-bit images contain pixels which have a value of either 0 or 1, thus utilizing the 1 bit of space available. Images which are created in 1-bit mode are called monochrome and each pixel is either black or white. This format is commonly used for line art. Fig.1, 2 and 3 below are an example of an image which has been converted into a 1-bit image and when zoomed in you can see that it is only made up from black and white pixels which in areas are formed in clusters to give the effect of shading/ gradation. Co Reg: VAT Reg: Reg Office: The AmBer Centre, Oldfield Road, Maidenhead, Berkshire, SL6 1TH, United Kingdom

2 (Figure 1) (Figure 2) (Figure 3)

3 8-bit (Greyscale) In order to represent enough tone levels so that the step from one tone to the next is not visible, about 200 levels are required. This should then give the impression of a continuous tone image. An 8-bit pixel can store a value between 0 and 255, giving a total of 256 different tone levels. This is calculated by using the mathematical calculation of 2 8 and so gives more tone levels than the required 200, but if seven bits were used (2 7 ) this would only give us 128 tone levels. These extra tone levels are actually quite useful, in fact, essential in the colour management world due to the inevitable loss of tones at each stage of production from the original digital capture, to the final printed output. Fig.4, 5 and 6 are an example of an 8-bit greyscale image. Note how much smoother the gradations are compared to the 1-bit image and when zoomed into pixel level you can see how each pixel is a shade of grey. (Figure 4) (Figure 5)

4 (Figure 6) 2-bit, 4-bit or 8-bit (Indexed Colour) 2-bit, 4-bit or 8-bit indexed images are used when a colour image is required, but file size is an issue. There are programs which take a true colour image and index the most commonly used colours to a palette which is saved within the file. This way each pixel has a reference into the colour pallet with a simple number, rather than storing the colour itself. It is also possible to create a new indexed colour image and select the colours you require from an existing pre-saved custom pallet, or a pre defined pallet, for example one which is optimized for the web, or for colours which the human eye has greater sensitivity. Commonly used indexed colour images are either 2-bit which can encode 4 colours (2 2 ), 4-bit which can encode 16 colours (2 4 ), or 8-bit which can encode 256 colours (2 8 ). The reduction in file size is an ideal format for multimedia or web graphics, but not suitable for full colour printing. Fig.7, 8 and 9 are examples of 2-bit, 4-bit and 8-bit indexed images, with their corresponding colour pallets alongside.

5 (Figure 7) (Figure 8)

6 (Figure 9) 24-bit (3 channel) 24-bit images contain 24-bits per pixel, which is actually 3 bytes. These images are classed as True Colour, meaning that unlike indexed images, they don t contain colour pallets and each pixel stores its own colour value. The confusing thing is that most people refer to these images as 8-bit, but what they are referring to is 8- bits per channel (Red, Green and Blue), which comes to a total of 24-bits. Using the formula 2 8 we know that it gives 256 tone levels, but now with 256 tone levels per channel combined, it gives over 16.7 million colours, or 16,777,216 (256x256x256) to be exact. Fig.10 below shows a 24-bit image. Note how much smoother the gradations are, giving the impression of continuous tone.

7 (Figure 10) 32-bit (4 channel) If a 24-bit RGB image is converted to CMYK, each pixel will require an additional 8-bits to store the black channel, increasing the image to 32-bit. This increase in bit depth will obviously increase the file size by one third. It is also worth noting that if a 24-bit image has an alpha channel added to it; this will require an additional 8-bits per pixel and increase the image to 32-bit. An alpha channel contains transparency information about an image and if saved with this channel to enable further editing, then it will remain as 32-bit. There are only a few graphical file formats which support alpha channels, for example TIFF, PSD, PNG and BMP. 48-bit With the latest high end digital cameras and scanners, it is possible to capture images in 48-bit. This means that each channel uses 16-bits (2-bytes) per pixel containing up to 65,536 tone levels, equating to 281 trillion colours (2 48 ). Considering that the human eye can perceive about 10 million colours and a 24- bit image contains up to 16.7 million colours, do we really need 281 trillion colours? It does seem quite excessive, until you consider the variety of conditions that we try to capture with our digital cameras, or sample with our scanners, but with these dramatically increased tone levels, it leaves a much greater buffer for when post processing of the captured images is required.

8 As an example, consider a photograph which was captured in 24-bit and later found to be very much under or overexposed. Instead of capturing the image with levels per channel, the camera may have only used 128 levels, resulting in an image being far too dark or light. Fig.11 and 12 below show a simulation of under and over exposure, along with their corresponding histograms, showing heavily weighted to the left, with nothing in the highlights when underexposed and heavily weighted to the right with no shadows when overexposed. The grey histogram in the background shows the correct exposure. (Figure 11) (Figure 12) In order to correct for this under or overexposure, the levels will need to be adjusted, redefining the white point and black point. This will then take these 128 levels and stretch them out to 0-255, but as there where only 128 tone levels in the original captured image it will be in steps of 0,2,4,6,8,10 etc and you now risk being able to see stepping (posterization) in your gradations.

9 With an image captured in 48-bit you have so many more tone levels that the same example would still leave thousands of tone levels and so still be able to give the impression of continuous tone in the gradients. In practice if images are captured with the correct exposure settings and post capture processing is kept to a minimum, then 24-bit capture will be more than adequate, but in the real world this isn t always the case, so if you have the ability to capture in 48-bit, then do so and then after making any exposure or colour corrections, convert the final edited image down to 24-bit, if file size is an issue. Summary Bit Depth Formula Colours Common Supported File Formats 1-Bit BMP, GIF, PSD, PNG, TIFF, RAW 2-Bit BMP, GIF, PSD, PNG, TIFF, RAW 3-Bit BMP, GIF, PSD, PNG, TIFF, RAW 4-Bit BMP, GIF, PSD, PNG, TIFF, RAW 5-Bit BMP, GIF, PSD, PNG, TIFF, RAW 6-Bit BMP, GIF, PSD, PNG, TIFF, RAW 7-Bit BMP, GIF, PSD, PNG, TIFF, RAW 8-Bit BMP, GIF, PSD, PNG, TIFF, RAW 24-Bit BMP, PSD, PNG, TIFF, RAW, JPEG 48-Bit PSD, PNG, TIFF, RAW