You might be pretty proud of that new camera you just bought and how it has a daunting megapixel count. Bigger enlargements are yours for the simple exchange of a few extra dollars right? Not necessarily.Many people ignore… Let me back up. Techniques seems to have suffered as a result of the digital camera revolution. With so many things repairable in software, photographers have become haphazard in their photographic technique. Image stabilization, color correction, and full auto settings, focus, etc on the in-camera side and powerful software tools on the editing side have let us slide into blissful laziness when it comes to taking a photograph.
I bet everyone out there has a photo like this one. You are perched on a hillside watching the dramatic sunrise looking at all those stunning colors, you snap the picture, review the LCD display and hoot (See Chimping) like a dominate ape. But when you get back to the home photo lab, aka your personal computer, you find that the image is slanted a few degrees to port or starboard. Continuing the naval metaphor, you decide to get your picture back on an even keel. Software can fix anything. Right?Every image editor out there has either a dedicated horizon straightening tool or at least the ability to rotate an image, either interactively or by degrees. Here is a quick tutorial I made for Corel PhotoPaint X3 users. So, you straighten that leaning image of Pisa, or that level water that appears to be running downhill to the right or to the left and all is well with the world. Right? Wrong!Some of your precious pixels leaked out. You didn’t even notice-did you? If you read Part 1 of my article on the three Cs of image editing will remember that straightening an image is the same as cropping an image.
As illustrated then, straightening a picture and yet maintaining that ‘parallel to the floor’ rectangular shape requires that little triangles of image be trimmed off of each side of your image. How much area is lost depends upon how far off from level your image is, and, to a certain extent, depends upon the format of your image. A camera sensor with a 3:2 aspect loses pixels at a different rate than a 4:3 sensor or a medium format with a square sensor. Since most DSLRs are 3:2, I did my calculations on that proportion. Though the rate varies, it is safe to say that you loose 3-4% of your total pixels in an image for every degree you are required to rotate to bring your image to the desired orientation. If you get to 5% rotation, the figure is a daunting 21.7%. (Pixel loss was measured using Corel Paint Shop Pro Photo Horizon straightening command and comparing image size before and after rotation of degrees on a 3:2 image.)
Now, if you paid the extra bucks to buy that Rebel XTi instead of the Rebel XT and you were negligent enough to let your camera get off by five degrees, your adjustment to straighten your image just cost you the difference in megapixels you paid to go up to the next model camera!
Beyond the pixel loss due to these cropped triangles, every pixel in your image suffers. Pixels are square. You can rotate them 90, 180, or 270 degrees without having to re-interpolate the image. If you rotate to a degree measurement other than the cardinal degrees, you are forcing your software to recalculate the values of every pixel in your image. You will lose detail. You will be farther from the reality you captured.
The Letter A in the example above is an illustration of non-cardinal degree rotations. The A on the left is as typed in the image editor: Corel PhotoPaint. The Center A was rotated 27°. The Right A has been rotated back 27°. Notice how much more sharp Left is than Right. This is a result of software interpolation of square pixels on rotation of less than 90°. Even a degree of rotation will introduce softness and loss of detail in your images.
The simple solution is to to take special care to orient your camera to the level items, be they vertical or horizontal, prior to snapping the shutter.
The bottom line is keep that bottom line straight. A really fine capture isn’t ruined by a slant but the ability to print it at maximum quality and size is diminished.
Rikk Flohr © 2008