A pixel is defined as: the smallest discrete element of an image or picture on a computer screen; a single element of a raster image.
Think of a tile mosaic where each individual colored tile is placed in an exact order next to other colored tiles to produce a larger image when viewed from a suitable distance. Each individual pixel represents one piece of color – and when placed together with enough other pixels, an image is formed, that to the human eye appears to be a continuous tone photograph. Whether scanned from film or captured digitally, all digital imaging today is reliant on these small pieces of color that come together to form a larger image.
How did they do it before computers?
Previous to modern digital imaging, photographic prints were made optically, and were not based on pixels, but were made using enlargers. On the publishing side, full color images were stripped in with color separated negatives – which were then shot with a prepress camera to produce cyan, magenta, yellow and black metal plates. Today most digital photographic prints are produced by sending digital images to various types of printers, the most common types being ink-jet and dye-sublimination. Photos being published in books or magazines are created by placing high-resolution image files in page layout applications, which are then color seperated into CMYK and output to imagesetters or platesetters.
How deep are your pixels? Understanding bit depth
Bit depth describes the number of bits used to represent each element in an image. A basic rule is that eight bits make up a byte, which will be important later.
Lets start with the most basic of all image types: line art. A true bitmap image of line-art is a single bit image, meaning there is either 100 percent black or white. This is expressed by two values only: 1 or 0, black or white, off or on. There are no shades of gray to a single bit image.
So each bit has two possible values – black or white. Eight (8) bits combine to make a byte – which is 2^8 (2x2x2x2x2x2x2x2), or 256 possible values or shades of gray. This is where the standard 256 level grayscale comes from.
A standard 8-bit RGB color image consists of three over-layed grayscale images, with one representing red, one green & one blue in various levels for each channel. Each channel has 256 possible values x 3 channels (256x256x256) which equal 16.7 million possible values. This is your normal RGB color image that is seen on websites or worked on in Photoshop from a digital camera or scan. The next step would be an image that is 12 or 16 bit, meaning that instead of eight bits per channel, there are 16 (2x2x2x2x2x2x2x2x2x2x2x2x2x2x2x2), which would give you 65,536 possible values for each channel. Instead of 16.7 million possible colors, 16-bit images can contain up to 281,474,976,710,656 (65,536 x 65,536 x 65,536) colors in a single exposure.
Now that you understand what pixels are and how they are made up, let’s move on to how many of them you need for various uses.
Previous to digital imaging, continuous tone prints were a much simpler concept – negatives and slides were enlarged to the needed size and a print was made. In today’s world of digital imaging, many of those without experience are confused by the relationship of incoming resolution (whether from a scan or direct from a digital camera) and output resolution to produce a digital print, a file for publication, or an image to display on a computer screen.
Note: the popular term dots per inch is actually not accurate when describing digital images that are not in print yet. Dots per inch (or DPI) would only describe the circular or elliptical dots on a printed photo or page. As the image exists in a digital file, they are pixels per inch. The more accurate terms PPI and pixels per inch will be used here instead of DPI or dots per inch when describing digital images not yet in print.
First, to achieve a print which appears to be continuous tone (ie: no visible dots), you need at least 200 ppi. Of course, this varies with each individual’s eyes and their ability to view these dots in print. But, as a general rule, 200-300 ppi is sufficient to create the appearance of a continuous tone print. Of course you want the highest resolution possible for any given output, up to about 400 ppi.
With the general rule of 200-300 ppi in mind, we can start looking at input and output resolution. The input resolution is the actual pixel size of an image. Regardless of the output size, the pixel size does not change. For instance, if you scan in a 2.25”x2.25” transparency at 1000 ppi, this will give you an image with a pixel size of 2250×2250. Keep in mind that a digital image has no absolute size or resolution. The number of pixels per inch is proportionate to the reproduction size in inches. At it’s scanned size of 1000 ppi, the image is 2.25”x2.25” at 1000ppi. At 500 ppi, the image would have an output size of 5.5” x 5.5” (1/2 the resolution = 2x the output size). Whenever you increase the output size, you decrease the resolution by the same percent.
Let’s take a real world look at this with a common file size, one from a 6MP digital SLR (DLSR).
When you open your image in Photoshop, you will adjust the output size and resolution by going to Image>Image Size. When you open the Image Size dialogue box, first make sure that ‘Resample Image’ is unchecked.
The native resolution of a 6MP image is 3072 pixels by 2048 pixels. At 72 ppi, this photo has an output size of 42” x 28” – and while that’s large, the resolution is far too low to produce a continuous tone print. When you enter a resolution of 200 ppi, the output size changes to approx. 10” x 15” – which is the largest output size you can have using the native resolution of the file. At 300 ppi, the file has an output size of 10” x 7” – which would be about as large as the file could go for traditional offset printing on a four-color press.
To reduce the image and the actual pixel size for web use, you are then resampling the image. To understand the best method for reducing the pixel size, lets first look at how the pixels are arranged within the file.
An exact grid is used to keep the pixels in order, which is determined by the native capture of the digital file. Preserving this grid is very important, as resampling the image on an odd multiple will force Photoshop to redraw the entire gird structure of the image, which can result in a “fuzzy” or degraded quality image.
If you have a grid that is 8 pixels wide by 8 pixels tall and you divide the grid by two, you’ll then have a grid that is now four pixels wide by four pixels tall, preserving the original structure of the grid.
But what happens if you divide the grid by 2.7? Then you will have a grid that is 2.692962962962962 pixels wide. The problem is there is no such thing as a half a pixel, or a 1/15th of a pixel – there is either a pixel or not. Photoshop is then forced to redraw the entire grid structure of the image, instead of just reducing the size of it. To preserve the original grid structure and achieve optimal image quality for web images, you must reduce you images by exact multiples of one-half (50%) or one-quarter (25%). Reducing a 6MP image to fit on a standard monitor at a reasonable size, you would reduce the image by 25% – which would give you a final pixel size of 768×512.
Photoshop CS5 uses advanced algorithms to resample images on odd multiples, which are designed to preserve image quality. However, many photographers and prepress technicians still believe the best method for resizing images is based on using exact multiples of one-half. Also, it is widely accepted to perform the resize in one single step, as opposed to multiple resizes on the same image. When decreasing an image in CS2 and higher, under the resample image box, use the option of “Bicubic Sharper.”
Along with reducing the image size by checking the resample image box, you can also increase the image size. When “ressing up” an image, care must be taken not to degrade the image quality to the point where pixelization will be noticeable.
As a general rule, a well-exposed, sharp 6MP image has enough raw pixel information to be increased by 200%. Using high quality lenses and a tripod will also help result in sharper images when resampling to a larger resolution. Any problems with the image, such as being slightly out of focus or poor lens quality, are going to be magnified as the size of the print increases.
At an output size of 8” x 12” a 6MP image has a resolution of 256 ppi, an acceptable resolution for a print of that size. But what if you want to produce a print twice that size? At it’s native resolution, a 6MP image would have a resolution of 128 ppi at 16” x 24” – not acceptable for a continuous tone print.
With the resample image box checked, choose “Percent” from the drop down menu that currently says “Inches.” Entering a value of 200 for the percent and “Bicubic Smoother” will now give you an image that is 16”x 24” at 256 ppi. After resizing, check the image at a 50% view and a 100% view. If the image appears too soft, then you can apply an Unsharp Mask filter to increase edge contrast and bring back some of the perceived detail.