This is a photo that had some discolouration damage, creases and scratches.
The finished image has had the discoloured areas removed, with a general overall repair, and enhance.
The simplest way to describe a pixel, or ‘Picture Element‘, is to define it as the smallest physical element in a digital image, or the most finite level of detail in the image. If you look at the image below, this is a zoomed in view to show some individual pixels:
Pixels are essentially like tiles, each one of a single hue of colour, and together they form an image when viewed from a distance. Each pixel can only have one colour value. So when you are looking at a digital image, such as in Adobe PhotoShop, what you’re seeing is an image composed on heaps of tiny coloured squares laid out in a grid. Pixel based images are referred to as raster images.
It’s not possible to get any further detail beyond this basic pixel, or to extract any further detail from within it.
If you watch those crime shows on TV, you’ll see some wonderful things happening with digital imaging technology that allow them to find hidden clues within the pixels of captured footage by zooming in and ‘enhancing’ individual pixels to reveal a word, or reflection of a person. It’s understandable that some people may think it is possible to do this to some degree. This is impossible. The image above is an eye and is very low resolution. The pixels above don’t hold anymore information to determine any further features of the eye whatsoever. This is a far as this image can go. To see further detail, you’d need to take a much higher resolution image.
I love this example below from a well known TV show of what can’t be done with pixels: http://youtu.be/IRBo5ZGcyVA
The number of pixels present within an image define the image’s pixel resolution. The final pixel resolution of an image is defined by the width expressed in pixels and height expressed in pixels.
Digital cameras capture images as pixel elements, and the more pixels per image, the higher the image resolution.
Megapixel ability is a term often used with digital cameras. A megapixel is about one million pixels. A camera that takes photos at three megapixels will have lower resolution images than one that can take images at 10 megapixels. This reference is to the resolution of the images, not the file size. In essence, the pixel resolution refers to the image quality, and the more pixels, the better the quality. Other mechanical factors with digital cameras can change this also – a cheap camera taking images at 10 megapixels will not capture images as well as a good quality 10 megapixel camera.
Resolution is also further determined by the required output use of the image and can take on more specific definitions. The above describes a finite image resolution in pixel size. Most designers will want their images supplied with a high pixel resolution.
A lot of people refer to 72dpi, or 300dpi images, when in fact they mean 72ppi, or 300ppi (which is pixels per inch). This is again an image resolution expression, but in this case this is just one of three measurements required for an image resolution for a specific output size and use.
When a designer or publisher requires an image of a certain resolution, at this stage, they are not concerned with pixels per inch. It really doesn’t matter as that comes later. Their concern is for an image that is of a good enough pixel resolution for use in various subsequent outputs, and will more than likely request the original, high pixel resolution image that is larger than the required output.
The reality is, when producing an image for a specific output size and use, the actual PPI of an image means absolutely nothing for resolution without a width and height attribute. It must have a resolution AND dimensions. Saying an image is 72ppi or 300ppi alone really doesn’t provide enough information as to the actual physical resolution required for it’s specific output. You can have a 300ppi image that is 100pixels by 100pixels, no good for print, or the same image that’s 72ppi but 3500pixels x 3500pixels, perfect for print. The designer will definitely want the 72ppi version in this case as all they’re interested in is the 3500pixel x 3500 pixel dimensions, and they’ll adjust the size and ppi where necessary as part of their mechanical processes for the required output.
An image produced by a digital camera can be adjusted to be any ppi (even to 1,000ppi), and it’s actual pixel resolution will not change in the slightest if its width and height pixel proportions are constrained (ie, kept at their original pixel width and height dimensions).
What PPI does do is determine the image’s physical resolution and dimensions when viewed at 100%.
Lets take a one megapixel image, and change the PPI, but keep the pixel width and height attributes the same. A one megapixel image is about one million pixels, which will have a width of 1,000 pixels and a height of 1,000 pixels. That’s actually quite a small image for print use.
OK, so can you just increase an image’s pixel dimensions to make a larger image?
Well, you can, sort of! Adobe Photoshop and other image editing software will allow you to do this, and they will fill in the ‘gaps’ between each pixel by a process of interpolation. However, at best they need to create new pixels from nothing to increase the image’s pixel density, and can only guess what sits between the existing pixels based on the existing pixels. Remember at the start I mentioned that the pixel is the most finite level of detail present in an image, so increasing the image size will not allow for more detail to be added that doesn’t exist in the original image. The lower resolution image can only capture a certain amount of detail with finite number of pixels it possesses.
This process can be used to ‘get away’ with some up-sizing to some degree – the quality is not that great, but visually, if you’re not too concerned, then it may do the trick. Ideally, as a designer myself, I usually avoid this at all times.
If you were to do the above, you would end up with something similar to this below. The left image is 79px x 100px. The right one has been interpolated upwards in Adobe Photoshop to double the width and height. At best, it looks blurry. it seems some detail has been regained, but this is in fact just an optical illusion. It just seems more detailed is there as are more pixels, but it will always be blurry:
Resize the original 79px x 100px image to 740px x 937px size as below, and this is what you get below. It will not get better, no matter how hard you try:
There are some interpolation software products online that use a myriad of calculation processes to try and guess what the missing pixels could be like, based on the surrounding pixels, and some might produce semi-ish OK results, but they are still guessing. There’s no way at all to regain information that isn’t there in the first place. Interpolation is in no way a substitute for not having an original high pixel resolution image.
In that case, if you do have a large image, can you decrease an image’s pixel dimensions to make a smaller image?
Images can size down to any size with no issue. Just make sure you keep the original, full resolution image. Don’t save over it. If you do, you can’t increase the size again, as image data, and therefore pixel information is lost during the downsizing process.
All pixels have what is known as a pixel depth, or bit depth, or colour depth.
The pixel bit depth is the number of ‘bits’ of information required to store the colour information for that pixel. In this post we will use Adobe Photoshop as an example.
When working with Abobe Photoshop, you generally work in 8-bits per channel. You can go to 16 or 32 bits, but most screens are 8-bits per channel. The more ‘bits’ of information per pixel, the more colours are possible.
The screenshot below shows this, with 8 Bits/Channel selected.
This means that there are 8 bits of information, per channel of colour (in this case RGB) required to store the colour information of every pixel. This will directly determine the file size of the image. What this means will become clearer in a moment.
A bit is the most basic unit of information in computing. Each bit can only have two possible values, which is usually 0 or 1.
8 bits equals one byte of information, and 1024 bytes equals one kilobyte (KB), and 1024 kilobytes equals one megabyte (MB), and 1024 megabytes equals one gigabyte (GB), and 1024 gigabytes equals one terabyte (TB), and this keeps going. These are binary multiples, whereas sometimes decimal multiples of 1,000 are used. So when you see an image file that is expressed as 6.24mb, that’s the file size in megabytes.
You’ve no doubt seen external hard drives for sale with their storage capacity expressed as 1, 2, or 4 terabytes and so on.
In Adobe Photoshop, and many other image editing software packages, images have colour channels. They are usually RGB (Red, Green, Blue) for digital use, or CMYK (Cyan, Magenta, Yellow, Black) for print use. There are a range other colour channels but they are beyond the scope of this post.
For this we’ll focus on RGB images
Bits per channel means that each channel, or colour value, for each pixel requires 8 bits of information per colour value to fully describe the colour of the pixel. So think of each pixel requiring 8 bits of information per colour channel with a combination of states of being on or off for each bit within each channel. You end up with each channel having two to the power of eight colours, or 28, which is 8×8, or a range 256 colour values within each channel.
So each pixel can have one of 256 values in the red channel, 256 in the blue channel, and 256 in the green channel, and these are express as 0 to 255 for each channel when mixing the three colour channels to produce a final RGB colour.
This vibrant green in the screenshot below has the following RGB values to describe it:
Red value: 152 / Green value: 232 / Blue value: 47
As a result of this, each pixel of the image requires 24 bits of information in total to describe it in full. 8 bits per channel, or 8 x 3.
Having 256 possible values in each channel will determine the total possible range of colours that can be achieved as well. 256 tones doesn’t sound like much! In fact, it’s a lot. Having three colour channels with 256 possible values in combination with each other. 8 x 3 results in a 24-bit colour image (24 bits per pixel). This is often referred to as a true colour display. The reality is you’re now looking at 224 possible colours, which is 16,777,216 possible tones of colour for each pixel. (That’s nearly 17 million). The other way to arrive at that number is 256 x 256 x 256.
This bit depth determines the file size (uncompressed), of an image when opened. For example, the RGB image below is 500pixels x 500pixels, and you can see at the bottom left of the image that the size is indicated as being 732.4K (kilobytes).
This is worked out as follows:
500 pixels x 500 pixels = 250,000 pixels (area)
250,000 x 24 bits per pixel = 6 million bits of information
6,000,000 divided by 8 = 750,000 bytes of information
750,000 divided by 1,024 = 732.421874 kilobytes (rounded down to 732.4kb in the image below). Neat!
So this image is just under three quarters of a megabyte when opened. (Please note, saving as .jpeg or .jpg applies compressions algorithms to the image that will reduce this file size when not opened. File formats will be covered in another post.)
This is a recent digital photo retouching of a black and white photo from the 1960s of a band playing a gig. The brief of this job was to restore some minor damage, and then colour the photo to approximately how the colours were at the time. I was briefed on the approximate colouring of the clothes worn by the band and their hair colour, the colouring of the drum to the left, and the general colouring of the surroundings and mood of lighting.
Especially important was the fact that the guitarist’s guitar had a sunburst finish, so it was important to get that to be as close as possible to how it would have looked.
Below are the before and after images. (Click on the image to have a closer look).
This photo below has been folded and creased and appears to be in a bad way! However, believe it or not, this photo is in fairly good condition.
When I looked at this photo, the first thing I noticed was that the girl’s face is intact, apart from the crease below the mouth, and the for the most part, there is no other real damage to the image apart from the creases themselves.
The creases were also running through areas that are mainly textures. Any features that are defined, such as eyes, nose, mouth and hands have escaped damage. As a result, this image was certainly repairable.
If you click the images below, you can see a larger version and can compare the before and after view.
If you have similar images like this, of a single subject against a general textured background, with the main distinguishing features intact (face and hands), and large areas of background intact, you’d be surprise to know that it can be repaired quite easily.