The magazine of the Melbourne PC User Group Looking inside the square! Laurie Weston westonl@melbpc.org.au

This article has been written mainly for those members who are considering using some form of digital photo manipulation on their PCs, or who wish to produce colour photos using their colour printers.

Current PC bitmap imaging, whether photo or video relies oil little squares known as pixels which exist everywhere except on the final output, whether we have an analog signal, a digital signal or lots of dots.

In this article, I will give you a brief summary of just what goes on inside those little squares and how they are able to work with each other to accept or display an image.

Monitors (Standard, not LCD etc.)

Colour monitors have lots of pixels, usually from 28-40 or store every centimetre. (72-100 or more pixels/inch). Inside each pixel are three phosphor dots red, green and blue, which are activated by three guns at the back of the tube. When red and green are activated, you see yellow (Y) in that pixel. Likewise green and blue combine to make cyan (C), and red and blue give magenta (M). These are the colours used together with black (K), in the CMYK printing process. Dependent on the monitor/graphics card combination, there can be 256 levels (8 bit colour) of each colour which allows for a combination of 16,777,216 colours (24-bit or truecolour). We won't go into bitplanes or lookup tables here, except to say that you look at 256 colours simultaneously, from a palette of over 16 million colours by using them.

Scanners

Scanners and other CCD (Charge-Coupled Devices) also have tlhese little squares. Lots more of them in fact.

Typically the number would range bctween 120 and 480 per centimentre in a consumer flatbed scannrer, and around 960 per centimetre on a consumer film scanner. Professional scanners have more because they are required to do a better job, while consumer units invent any extra squares they may need. This is known as Interpolation.

Each pixel in a CCD array responds to the colour of light it receives with one or more bits of information.

The bit depth of a consumer device is usually either 24-bit or 32-bit. As with monitors, 24-bit refers to 256 shades of each colour. The remaining 8 bits in 32-bit colour are used for what is known as an Alpha Channel which describes a transparency value for the pixel. When an image is later separated into distinct CMYK channels, 32-bit world then refer to 8-bits per each of the four channels. Separation, however, is not in the scope of this article.

Some scanners and image processing software also support 48- and 64-bit colour. Scanning at such high bit depths causes both high colour fidelity and huge files with more information than most software can handle. The scanning software usually analyses the file before saving and reduces it to a 24-bit subset of the most commonly used colours.

A word of caution: Scanner resolution is usually quoted in dpi (dots per inch). That is inaccurate; it should be ppi (pixels per inch) and is the first of two common misconceptions that a salesperson may have.

The second misconception: a scanner that has a resolution of 1200 x 600 dpi is optically a 600 dpi scanner, not a 120 dpi scanner. The higher figure indicates that there are two banks of 600 ppi resolution CCDs.

I have lost count of the number of salespeople who have tried to tell me the higher value is the scanner''s optical resolution. Many often quote the highest interpolated resolution. Don't ever consider interpolation when purchasing a scanner. The image information its an optical pixel is the same as in a group of four interpolated pixels at 4800 dpi.

Camcorders & Digital Cameras

Camcorders and Digital Cameras have a fixed number of pixels. In PAL video systems, the grid is 740 x 578 pixels. They also invent higher magnification by actually increasing the displayed size of those pixels. Otherwise referred to as Digtal Zoom. Digital cameras vary froth 740 x 578 upwards (early models used Camcorder chips). Manufacturers currently try to blind you with science by using the term mega-pixel. That only means more than one million pixels (total) exist on the chip. A matrix of 1280 x 1024 pixels will have 1,310,720 pixels, which would place it in the mega-pixel category. We will shortly discover why you don't expect to blow up a poster print from that image.

Squares and Dots

To correspond with these pixels, newt printers use dots of various resolutions. While there is similarity in the resolution of these dots to the resolution of pixels, there is not uniformity. Printers have the ability to increase the size of the dots to various degrees, according to such factors as the type of paper being used. Pixels always stay the same size; at least its theory. For this reason common usage has done us a disservice in referring to all forms of resolution as dpi. This will become store evident when we move to different forms of both scanning and printing.

In determining the relationship of the number of pixels required in the source image, to give the required number of dots the the printer to use, there are many factors involved. Much of this relationship can he worked out only by trying different settings with your printer and discovering what works best in terms or how you want your picture to look. These tests are part of setting up all future projects and will be referred to later.

I offer the suggestion that although modern inkjet printers have a 1440dpi capability, most people seem to use something more like 300dpi for the output resolution when determining the scanning resolution. This will be covered shortly.

Putting it all together

How do we use these little squares in electronic bitmap imaging?

Photographs

The characteristics of a photograph are described in terms of image size and magnification. Higher resolution is achieved by using larger negative sizes. Less grain at a given magnification is achieved by using finer grained emulsions. In short, conventional photographers are concerned with the area of a negative and the area that will be covered in the final print. This determines both the type and size of film used and the focal length of the lenses.

Digital Photographs

Digital photographs are worked in terms of pixel resolution and printed in terms of printer resolution. Digital images are more about the number of pixels required on the longest dimension of the image. The rest will fall into place. The relationship between the electronic image and the hard copy image is the first and most important factor to consider at the start of a project. Yes, I said project because the whole digital process has to be considered as one, whereas in conventional photography there are two distinct stages, exposing film then printing, each of which can be variable.

Project 1

Given a photograph on film, negative or transparency, produce an inkjet print of a given size.

Factor 1: How many pixels are required along the longest dimension of the image that I am sending to the printer?
Factor 2: What is the corresponding length of my source image?
Factor 3: What resolution do I have to scan the source image, to give me the required number of pixels?

Fortunately, the majority of better image software packages do the mathematics for you; all you need tell the program is the output printer resolution in dpi and the software will work out the scan resolution. The resolution you choose is not necessarily the resolution you will actually set for the printer, but rather the resolution you have worked out by your tests mentioned previously. For this reason I will refer to an output resolution as ppi because the resolution I will be referring to is part of the file, not the output. I'll give you a practical sample, but you have to work out what is best for your own situation.

Practical Example

Working with the commonly used imperial standard, I am to print an A4 size image from a slightly cropped 35mm slide.

I choose an output file resolution of 185ppi. Why was 185ppi chosen? Because from various tests it was determined that using a higher output resolution used more ink, when converted to dots by the printer, but produced little extra quality. Conversely, using lower output resolution showed an unacceptable print quality with this particular printer. Length of image: allowing for border will be 10 inches. I will require 1850 pixels on the longest side of my image. Length of source used is 1.3 inches My scan resolution will be 1423ppi. My printer will be set at 1440dpi
Result: My image will be 10 x 7.7 inches on an A4 sheet

If the print were to be produced commercially by an Iris or Pegasus printing system a set resolution would be that which the Bureau recommended and could have been eg 400ppi. Of course in this instance they would tell you to use 400dpi.

Project 2

I wish to print from a single frame grab from a video camera.

Factor 1: I already know the number of pixels on the longest dimension of my image is 740 pixels.
Factor 2: Choosing the same output resolution as for the previous sample project (185ppi), my maximum length on the output image will be 4 inches.
Result: My image will be 4 x 3 inches (approx.)

Project 3

I wish to print from a digital camera image 1280 x 1024 pixels.

Factor 1: I know already that the number of pixels on the longest dimension of my image is 1280 Pixels.
Factor 2: Choosing the same output resolution of 185ppi as for the previous projects my maximum length on the output image will be 7 inches.
Result: My final image will be around 7 x 5.5 inches.

Project 4

I wish to print an A4 print from a Kodak Photo CD.

Factor 1: Find the image size to open the pcd file that best suits the print size required. Poster size is 3072 pixels on the longest dimension. Large size is 1536 pixels on the longest dimension. Standard size is 768 pixels on the longest dimension. Snapshot and Wallet sizes are too small for consideration.
Factor 2: At 185ppi what are the lengths of the print from the three given Kodak image sizes? Poster size would be 16.6 inches long. Large size would be 8.3 inches long. Standard size would be a shade over 4 inches long.
Factor 3: Decide if you want a print that has approximately 1.5 inch border or one that fills the page. This is because Poster size would be too large for the page and large size would be slightly too small to fill the page. Of course a third option would be to print four standard size images on the page.
Factor 4: Assuming the full page (less small margin) is required the choice would have to be, to open at poster size and then re-sample down to the required number of pixels using the imaging software. Alternatively you could open at large size and resample upwards if you were prepared to accept the slight loss of quality.

Summary

An understanding of pixel resolution, bit depth, dot gain and other associated factors is necessary if you wish to have high quality results in computer manipulation of photographs.

While we are still using these little squares for bitmap images, it is necessary to look inside the square.

Reprinted from the October 1999 issue of PC Update, the magazine of Melbourne PC User Group, Australia

[About Melbourne PC User Group]