As others in this thread have commented, that depends a LOT on the image – what it contains, what format you set it to export as, a hundred other nitty details.
tl;dr: More than you wanted to know about image compression, Freeway’s graphics engine, and the trade-offs between flexibility and quality.
###Compression
Most image compression schemes use deliberate data loss as part of their size-reduction methodology. JPEG uses a “supercell” approach, which looks for the largest block of image data that can average to a single value or repeatable pattern. Then it also adds some artificial “sharpness” to boundaries with high contrast.
Between those two basic tricks, it can often represent an image with far less data size and no apparent loss of detail. The problem is, these tricks are best used on an image like a landscape or a portrait against a woodsy background – something that doesn’t have stark areas of flat color or geometric shapes. Those sharp edges and flat colors bring out the worst in the JPEG algorithm, exposing its tricks in an all-too-visible form, or (if you’ve set the “quality” slider higher to compensate) not really giving you any measurable reduction in data size.
GIF uses a different approach, known as run-length encoding. In its simplest form, it processes the image line-by-line, looking for repetition. When it comes across >2 pixels in a row that are the same color, it doesn’t store each pixel, just the start and length of repetition. (This is how fax machines work, and why they can send a 300ppi or 600ppi bitmap over an ordinary voice phone connection in a meaningful length of time.) This is an over-simplified explanation of how the encoding works, but it’s a useful abstraction when you’re learning about compression. Because it is skewed so strongly to large areas of flat color, and because if you use it on an image with fewer than 256 total different colors, it is effectively a lossless compression scheme, GIF is a perfect format for any form of art that is made up of flat colors, geometric shapes, even (with some tricks) gradients of color, or other things that JPEG chokes and dies horribly while processing.
PNG comes in two flavors – 8-bit and 24-bit. The former is very similar to GIF, but works around the patents (now expired) that kept GIF from being a truly open and free image format. 24-bit PNG (more accurately, 32-bit PNG) is a full-color format (>255 colors) comprised of three 8-bit image channels with an 8-bit alpha channel for semi-transparency, soft-edged masks, and other fine details. 8-bit PNG can compress very tightly, just like GIF, when the source image has a limited palette of colors, and 24-bit PNG does not compress very much at all, it just looks as good as it can, and does things no other format can do.
###Freeway
Now the next thing to consider is Freeway itself. The graphics engine in Freeway can do many amazing things, and in Freeway Pro, can do these to nearly every image format that Photoshop can open and save. You can crop, resample, rotate, filter, add effects, borders, partial transparency, just about anything at all. If you layer (and combine) a JPEG and an animated GIF, you will get a very large animated GIF that still works. This engine runs on any image you import into Freeway that you DON’T set as pass-through, and it happens every time you publish with changes to an image. In order to have the publish process complete some time during this epoch, Freeway uses a technique known as bilinear interpolation to create the “in-between” pixels that are computed each time you resample an original image to a different-sized output image. This technique is in stark contrast to Photoshop’s (and other high-end image applications’) bicubic interpolation. Because bilinear interpolation only operates on a fraction of the same amount of data, it doesn’t have as many clues to work with when computing the “fake” output pixels, and so the results are not as accurate. This is particularly evident if you are working on an image containing very slightly diagonal lines (which will stair step like mad) or sharp boundaries of color, like logos and type.
Long ago, like maybe in Freeway 2 (pleistocene era), there was a brief experiment in enabling bicubic interpolation. It was so slow to publish that beta testers thought their computers had crashed. (Not an unreasonable assumption; this was under Mac OS 8.) As far as I know, it hasn’t been tried again since, although with the incredible speed and resilience of modern hardware and OS X, maybe it should be.
###Workflow, and Flexibility
I’ve given this advice before, and will continue to do so until something changes. Treat Freeway as a sketching tool in the first place. Make your layout look the way you want it to, then go back over it and “sweeten” things until it’s perfect.
Maybe it’s my heritage, going back to using nasty Xerox copies of original photos as FPO images in my actually-glued-together layouts and mechanicals, continuing on to using QuarkXPress with its truly unreadable pixelated proxy images for the same purpose, but I have learned to overlook the fine details in a layout, focusing instead on the shapes and the rhythm of the type and the images first and foremost. (This probably also harkens back to my days as a studio photographer. Looking at something upside down and backwards, under the sweaty shade of a dark cloth forces you to look at things in an abstract manner, and can lead to better composition in my experience.)
Pass-through images give you the most control over image quality, but you throw away every other advantage that Freeway gives you in exchange. So use them, but use them LAST. Get your layout looking the way you want to. Squint your eyes a little so you don’t think about the jaggies. Take a screenshot and rotate it 180° to simulate the view camera (I swear: this really does something for your layout!). Show it to your client, get buy-off on the layout. (It helps a lot if you’re using stock photos at this point, because you can leave the nasty X-and-a-logo watermark in place to remind them that the photos aren’t “final”.) Then go back over the layout with your magnifier. Find the images that Freeway is not kind to. Follow these steps:
- Control-click on each image and choose Export… from the contextual submenu. Give it a meaningful name and file-type extension. This will be your “FPO” (For Position Only) image.
- Open that image in Photoshop or the local equivalent. Open the original, drag it into the Freeway FPO image as a new layer, and set that layer to partial opacity, so you can see through it to the original below.
3, Scale and crop the full-quality layer until it matches the pixels below it, then raise the opacity back to full.
- Export as a JPEG or PNG or GIF, using all the bells and whistles you have at your disposal. Photoshop is unbeatable at this: there are loads of controls in the Save for Web panel.
- Back in Freeway, click once on your starting image box, press Command-B to clear the image, then Command-E to import your “sweetened” image, and be sure to choose Pass-through when you do.
- Repeat as necessary. Remember, you don’t need to do this to everything (unless you really want to). You should look critically at the results of Freeway’s native image processing, because for many things, you can’t do better in another application.
Walter
On Feb 13, 2014, at 9:23 AM, Carla wrote:
Thanks guys… well at least i know I can use html box for an image as I have rarely done this… Although it is still odd the image looks so bad uploading in the graphic box, but oh well thx!
freewaytalk mailing list
email@hidden
Update your subscriptions at:
http://freewaytalk.net/person/options
freewaytalk mailing list
email@hidden
Update your subscriptions at:
http://freewaytalk.net/person/options