File Size Basics
There are three primary ways of saying how large a digital image is:
- the total pixels recorded at the time of exposure, usually in megapixels (MP)
- the amount of disk space a file takes up, typically measured in kilobytes (KB) or megabytes (MB)
- the number of pixels wide by number of pixels high (W x H) displayed on screen
The following is a three-step process used to convert a large original to an image small enough for e-mail and downloading, yet still retaining essential content. Until everyone around the world is on a high bandwidth connection, limit your images for email attachments and web posts to ones that are less than about 1000 X 800 pixels and less than 300 KB in size. And even those are larger than you normally want to deal with. The steps to transforming your 6-12MP originals to user friendly jpegs are:
Cropping is the first step in reducing the size of any image. Cropping selects a portion of an image and discards the rest. Cropping is the first step because it has no effect at all on image quality. Crop to eliminate non essential or distracting material from the edges of the image.
Find the cropping tool in your image processing application and use it to make a selection from the image, tightening up the original to the most critical portion and eliminating the non essential surrounding area. But don't go so far that you eliminate all context for your subject as that can make an image too sterile and boring in my opinion. You are setting the overall composition in this step.
The crop tool has been selected from the tools menu on the left side and used to draw a crop box on the image. The area that will be discarded is darkened; the crop box can be adjusted before committing to the actual crop.
Once extraneous material has been eliminated by cropping, the overall dimensions of the image should be reduced by resizing. An image that is 4000 x 3000 pixel W x H can be reduced in size to 1000 x 750 pixels by discarding 3 of every 4 pixels in every horizontal and every vertical line. This step is essential to reduce the image to a size that is comfortable to view on the web. To preserve image quality, crop as much as possible first. The image quality will be reduced by this step and the image may get a little softer and require some sharpening.
Select Image Size or Resize from the menus in your application and change the long dimension. Note that you want to keep the proportions the same to avoid a squeezed distortion. Choose "bicubic" if it is offered. Here's the panel in PS CS3 downsizing an image to 500 pixels wide:
Note: you can use this same tool to increase the dimensions of an image. For example, if you crop a distant photo to 300 x 175 pixels it will look much better posted at that size than if increased to 1000 x 800. Increasing the number of pixels adds interpolated pixels which almost always degrades the image. so this is not a technique that I recommend as a normal process.
The final step in reducing the size of an image for web viewing is setting jpeg compression. This is a visual "trick" that discards subtle information in the image to yield files that take much less physical storage (KB). Images can be easily compressed by 10:1 or so without any significant visual loss to image detail.
Compression comes into play when you perform a "Save As" action on the file. Select "jpeg" as the file type from the Save As panel, and then adjust the quality slider to get a file size in the range of 100-200 KB. Note that I always embed a color profile - it is essential that you do so if you want your image to look the way you want in a web browser that is color aware (like Safari and Firefox on Macs).
Name your file and it will be ready to send as an e-mail attachment that won't cause lengthy download delays even over dial-up connections.
PPI is an independent parameter that has nothing to do file size or with image quality. What PPI does is instruct the printer on how to allocate the total pixels in an image to cover the print. PPI tells you how big of a print you can make from an image at a certain print quality: a low value PPI spreads the pixels too thin and makes the image look blocky; a PPI larger than a printer requires (usually in the 240-360 ppi range for consumer inkjet printers) produces an image smaller than it needs to be (i.e., it "wastes" pixels).
For an image displayed on the web, PPI is completely meaningless. All that matters is the total pixel dimensions of the image: pixels wide by pixels high. An image of fixed width and height will look exactly the same when posted on the web if the PPI is 10 or if the PPI is 1000. Computers ignore PPI in an image when sending it to the monitor. The native PPI of the monitor determines how many inches wide and high an image will be; not the PPI embedded in the jpeg file.
Some of the confusion surrounding PPI comes from image processing software when users don't pay attention to what they are doing. In photoshop, for instance, you ned to turn off Resampling in the Image Size dialog if all you intend to do is change the PPI value; with Resampling On photoshop will hold the Inches wide and high values constant and toss out pixels (resample) if you change to a lower value PPI.
Example: Summer Tanager, Boyce Thompson Arboretum.
Image width: 720 pixels
Image height: 546 pixels
File Size: 134 KB
Jpeg compression: 1:11
PPI Top Image 100
PPI Bottom Image 400
PPI Top Image 100
PPI Bottom Image 400
They look the same size because they are the same size: PPI has no role in monitor display or file size or image quality.
My Typical Work Flow
- Remove the compact flash card from the camera and transfer the images to my Mac with a card reader. I do not delete the raw files from the CF card until it is reformatted in the camera the next time it is used in case something goes wrong with the file transfer.
- Review these raw files and discard the obviously bad ones - those with missed focus, grossly wrong exposure, subjects blurred from their motion, bad compositions, etc.
- Name and number the keepers - for birds I use the species name and a sequential number.
- Before any image processing occurs, all these images are backed up to an external drive and also burned to CD.
- open the raw file in Adobe Camera Raw (ACR,) make the initial adjustments on white balance and exposure
- pass the image to Photoshop by clicking the open button
- examine every part of the image for dust bunnies at 100% and remove them with the healing brush or clone stamp
- make adjustments with levels, curves, and saturation using layers (sometimes on selected parts of the image)
- save the full sized layered .psd file for future tweaks
- flatten the image and set to 8 bit
- crop the image to the composition I want
- save the cropped file as a .tif
- resize as needed for use (720 x whatever for web posting) and save a copy with a new name
- duplicate the layer to do selective sharpening with Smart Sharpen
- evaluate the sharpening by toggling the layer on/off and adjusting the intensity if neccessary
- flatten and save
- add signature/copyright from a separate file, drag it around and set a color
- flatten and save
- save a copy as jpeg
Shooting in Raw
Yes, these are all valid issues, but the benefits of shooting raw are so strong that not using this format seems like a waste of a good camera.
Raw format retains all the information captured by the photo sensors in the camera, without applying all the steps required to turn the raw data into jpeg images. Instead of having the limited processing power of the camera apply the shooting parameters set by the photographer from the camera menu before the shot is ever taken to get jpeg output, all the raw data is transfered to the photographer's desktop or laptop computer so that each image can be treated independently while the photographer judges each choice of what parameters to apply. If you care about getting the best results from your images then shooting raw is a necessity.
I use Adobe Camera Raw (ACR) to do my raw conversion when I import a raw file into Photoshop. ACR lets me make adjustments to white balance: often necessary even though the Auto White Balance setting in my Nikon cameras does a good job of determining the color temperature of the light when I take my photos. And if you accidently set the camera's White Balance incorrectly (like "tungsten" when outside), it is easy to reset it correctly in ACR.
Raw conversion in ACR also allows for a surprising degree of exposure correction, so those over or under exposed shots can be adjusted during conversion with tolerable amounts of loss to image quality.
The adjustment sliders in ACR can also help a great deal in fitting the dynamic range of the scene into the limits of the photograph. ACR displays a histogram showing how the tones of the image are distributed from complete black to complete white. Any part of the range that lies outside the black and white points shows up as spikes at the left and right ends of the histogram; this information will be lost when the raw file is converted and imported into Photoshop.
Five sliders control this histogram: Exposure, Recovery, Fill Light, Blacks, Brightness. By adjusting these sliders the image data can be made to better fit in the boundaries of the histogram:
- Exposure will pull the entire graph left or right
- Recovery will pull the highlights out of the "stack" at the right edge into the histogram range
- Fill Light will move the data from the left end of the graph away from the darkest point
- Blacks will move the black clipping point into the histogram
- Brightness moves the middle of the histogram
These two screenshots from ACR show how a histogram that "spills out" of the range on both sides can be brought mostly within the limits by adjusting these five sliders. Note there is still a small amount of shadow detail lost after the sliders have been adjusted (the narrow blue spike on the left of the "After" histogram); I'd rather lose a little here than on the highlight side.
Calculating Crop Percentages
But there's no obvious standard way to specify how much an image has been cropped, and a lot of inaccurate and confusing comments are made trying to indicate this. To me, it's simple: express pixels remaining after the crop as a percentage of the pixels before the crop. Let's take a 6MP original and crop it as an example:
Original: 3000 px wide by 2000 px high = 3000*2000 = 6MP
Crop: 2500 px wide by 1500 px high = 2500*1500 = 3.75MP
So, the cropped image is (3.75/6) = 62.5% of the original.
Two simple multiplications, and one division, and you've got the answer.
Note: you can set up Photoshop to show the pixel dimensions of any open image in the status bar at the bottom of the window, so it is easy to read the before and after dimensions of the image when you crop it.
This Redhead example shows a full frame image, which is labeled 100% on the outside border. The Yellow border shows 75% of the image (25% cropped away outside the yellow). The Orange border shows 50% of the image (50% cropped away). The White border shows 25% of the image (75% cropped away).
Shutter Count from EXIF in Photoshop
Baseline Your Gear
Before jumping in to the question of why an image taken with a long lens isn't as sharp as expected I think it is fundamental to establish the baseline performance of the gear - you really need to see what even a modest DSLR is capable of delivering to see how far off the mark any other image might be.
Find the modest "kit" lens that you got with your first digital body, like the 18-70 that came with my Nikon D70. Set the camera to the lowest ISO value (usually 100 or 200). Put the camera on a solid tripod and use it t take some shots of a well detailed stationary subject 15-20 feet away in good directional light. Stop down the lens a couple of stops from wide open when you do this. Shoot in raw mode (or at the highest quality setting if you still haven't given up shooting jpegs).
Now look at the results and see how much detail your camera can deliver under the best conditions. If you can't get a sharp image doing this, then there's a fundamental problem with the camera, the lens, or your basic understanding of how to use the camera.
Now hang on to these images and use them as a yardstick as you start to make the compromises needed to photograph tiny active birds in the wild. Compare these baseline images to those at higher ISO, or hand held, or with that "bargain" telephoto zoom and converter. By doing each of these comparisons while keeping the over parameters constant you'll have a much better chance at figuring out just where the big losses in IQ come from and then you can work on eliminating them.
Sharpening in photoshop is really just increasing the contrast between adjacent pixels: the transition across neighboring pixels looks sharper if there's more contrast between them. The sharpening tools in Photoshop (like Unsharp Mask and Smart Sharpen) work by increasing local contrast - the radius slider determines the range of pixels to work with; the amount slider controls the strength of the change in contrast over these pixels.
The real skill in sharpening is learning how to control the tools and applying sharpening to only the areas that need it. Oversharpening is a common mistake, and is often more of a problem than not enough sharpening. An oversharpened image usually looks "crunchy" or "scratchy," with edges showing too much local contrast. When the subject is against a uniform background (like a bird against a plain sky) it is easy to get a "halo" where the sharpening causes a lighter rim to appear in the sky immediately next to the subject.
Sharpening should be the final process applied to an image - do all the color and brightness adjustments first on the full size image, crop, and resize before doing sharpening. The resize process is one step that can result in the dreaded halo of oversharpening if the wrong option is selected. The resize panel in Photoshop is misleading as it suggests that you should select Bicubic Sharper when downsizing, as seen in the following screen shot
Yes, this can be a quick and convenient way to "punch up" an image, especially when it seems a little less sharp than it should be. If you only want a quick downsize and don't care about controlling the sharpening then this is a handy option.
But if you want to control where sharpening occurs and how much is applied, then avoid Bicubic Sharper and instead stick with simple Bicubic as shown in this screen shot
And if you do more than one round of downsizing (say you want both a 1200 pixel and a 720 pixel version of the image) then you definitely want to avoid two rounds of Bicubic Sharper which will over emphasize the edge sharpening effect and quickly lead to a halo.
So, lets start with a 720 pixel image that has had no sharpening applied, and was downsized in two steps using simple Bicubic resampling:
It shows good detail throughout the towhee's feathers, but shows the overall softness typical of digital capture.
We could punch up the image by using Bicubic Sharper for two rounds of downsizing, and then apply Smart Sharpen to the entire image for this result:
For me this looks oversharpened, with the feathers starting to look crunchy and a bit of a dark halo forming around the head.
I prefer to use Smart Sharpen and to apply it selectively on a separate duplicate layer. By working on a separate layer I get even more control as I can adjust the layer mask to control the selection, and I can reduce the sharpening by lowering the layer opacity.
Here's the unsharpened image with my rough selection indicated by the ruby red Quick Mask.
I use the Quick Select tool to make my initial selection - this tool is amazing on subjects with well defined edges, and gives a good starting point for subjects with less well defined edges. I switch to Quick Mask mode when needed and make corrections to the selection with black and white brushes.
Here's my final version of this Abert's Towhee image, using simple Bicubic sampling for down sizing and Smart Sharpen applied only to the towhee selection: