trying to get the best quality export to a pdf... still gets pixelated. Newbie

[pauland]

Thanks Acorn. It's interesting, but I'm not a believer.

[mwenz]

In the ZIP file are both the original JPG and the resized one. The original has pixel dimensions 576 x 432. The resized version has the pixel dimensions of 2304 x 1728 (400% increase).

A better beginning image produces a better result. This was done using SmillaEnlarger. I have used it quite a bit to save my bacon, though I wouldn't generally go 400% unless I had to.


Mike

[JET]

Anyone have any laymans explanation to what is happening when you export and use settings to set dpi.

The oft-cited "rule" that everything destined for print must be at least 300 PPI is a gross misconception due to a corruption of a rule-of-thumb dating back to the beginning of desktop publishing.

The rule-of-thumb is not "every image has to be 300 PPI (Pixels Per Inch)." It is that the effective PPI (Pixels Per Inch, at final scale) of raster images should be 1 to 2 times the halftone screen ruling. So-called "magazine quality" offset printing commonly uses a halftone ruling of 150 LPI (Lines Per Inch). Those not really understanding the matter corrupted the rule-of-thumb into 300 PPI, assuming that with the prescribed range being 150-to-300, 300 must always be "best."

Not so. First, the rule-of-thumb is a range from minimal to maximum. If one is going to assume an "ideal" in that range, it would be 225, not 300.

Second, the whole point of the matter is to ensure that pixelation (so-called "jaggies") does not occur in the final print. Pixelation is due to image pixels (which are square) being scaled so large that their square shape becomes renderable by the haltone dots. If the pixels are larger than the halftone dots used to render them, then it's at that point that pixelation becomes possible (thus, the low end of the "acceptable" range being 1:1). The upper end of the range is just to provide a measure of leeway for on-the-page enlargement, after importing the raster image into the page-assembly program.

Third, excessive PPI can actually degrade printed results, because all the excess data gets averaged away. Averaging colors tend to dull them, just as mixing adjacent areas of bright colors of paint on a palette dulls their brightness. Similarly, the unavoidable averaging of oversampled images tends to blur them.

Don't believe it? Assuming you sample your images to the appropriate PPI before importing, and do not perform on-page scaling thereafter; and assuming you understand that practically all scanned images (that includes digital camera captures) need some sharpening and know how to do that--you can test all this yourself on your next press run. Just position a few instances of the same photo in the waste area of the press sheet. Starting with the same original, create three images: one sampled to a PPI equal to the halftone ruling, one sampled to 1.5 x the halftone ruling, one sampled to 2 x the halftone ruling. Properly color-correct and sharpen each. After printing, separate them. Without mentioning their resolution, get an array of people to tell you which of the three appears the sharpest and has the "cleanest" color.

I am learning that there are several different dpi types (print, page, image...).

There are. And that's why, for clarity, it's highly advisable to adopt the habit of using less ambiguous terms. Instead of referring to everything as "dots per inch" as many do, use:

PPI (Pixels Per Inch) when referring to the effective resolution of a raster image. Understand this for what it is: It's really just a scaling factor. The raster image has a fixed number of pixels. That fixed number of pixels (the pixel count) is spread across a given linear measure, determined by the scaling of the raster image on the page layout. Also understand: PPI is for print. It's pretty much meaningless for on-screen (internet browser) purpose. When preparing images for the web, just use the actual pixel count (so-many pixels horizontal by so-many pixels vertical).

Not all images on the page must necessarily have the same PPI. Much depends on the nature of the individual images. For example, assuming a halftone ruling of 150 LPI, your properly color-corrected and sharpened photos will likely be in the vicinity of 225 PPI. But there is no reason whatsoever to insist on the same PPI for raster images that are nothing more than special effects, the most common example being the ubiquitous soft-edged drop shadow. The whole purpose of a soft drop shadow is to be blurred. So you only need the minimum PPI for a drop shadow sufficient to prevent pixelation. I'm sorry, but there's just no way a round halftone dot measuring 1/150th of an inch is going to be able to render the "squareness" of a pixel the same size.

Another example is the frequent (and frequently downright comical) fretting of some designers over the PPI of computer screenshots in things like software instructions. I've actually seen designers take captures of the screens they are depicting, open them in Photoshop, and then upsample them to the "everything for print minimum" of 300 PPI--to absolutely zero benefit. The entire intent of a computer monitor image is to show what it looks like on the monitor.

On the other hand, some raster images on the page may not be halftoned at all. 1-Bit images (as one might use, for example as the "inking" line art of a comic strip) are commonly rasterized to 1200 PPI, because there will be no halftoning to effectively mask their stair-stepped "jaggies" like there will be in other higher bit-depth images on the same page.

LPI (Lines Per Inch) when referring to the halftone ruling at which the piece will be printed. Understand, this is the number of rows (lines) of halftone dots spanning an inch. This will be the same value for the entire page, regardless of whether individual images on the page have different PPI. This is why, for most color images, the appropriate PPI is related to and based upon the halftone LPI, not to the printer's hardware resolution.

SPI (Spots Per Inch) when referring to printer spots; the actual size of the spots which the imaging device (be it a PostScript imagesetter or a desktop laser printer) can physically print. This is the actual "hardware resolution" of the printing/imaging device. In conventional halftoning, each halftone dot (LPI) is built out of printer spots. So the number of different sizes of halftone dots (levels of gray) a given output device can produce at a given halftone frequency (LPI) is directly dependent upon its SPI. If a color separation of an image has, for example, 256 different tones, but the printer only has enough printer spots to make 100 different sizes of halftone dots, then the halftoned image is going to become posterized (common on desktop laser printers or "digital presses" like Docutec devices).

Also understand, LPI refers to conventional halftone screening, which is still the most common screening method for offset lithography. It really doesn't as directly apply to stochastic screening, which is most commonly used on desktop inkjet printers and one-off large format composite printers. Basically, conventional halfoning uses uniformly-spaced halftone dots which vary in size (amplitude) to simulate varying tones. Stochastic printing does the opposite: It uses uniformly-sized dots which vary in spacing (frequency) to simulate varying tones. Thus, conventional halftoning is often called AM screening; and stochastic, FM screening.

Output devices which render stochastic toning span a "high end" to "low end" spectrum. Most composite (full color) devices fall more toward the "low end". That is, their resolution is not considered high within the realm of stochastic printing and one of the reasons they employ it is that it enables relatively low-res devices to achieve the "normal standard" capability of 256 levels of gray despite their relatively lower SPI. Examples include devices ranging from cheap desktop inkjet printers to expensive large format devices used for point-of-purchase displays and large signage like trade show displays and even billboards. One of the inherent advantages of the lower-end stochastic method is that it is more forgiving of lower image PPI. So this is another common situation in which the overblown 300 PPI "rule" is bogus.

Viewing distance is also a factor. Even if you insist on clinging to the mythical "necessity" of 300 PPI for magazine quality offset, you certainly don't need it for most large format printing. Whether you're talking about posters (other than very high end "fine art" prints), trade show displays, or billboards, you're talking about graphics which are not meant to viewed at 14 inches from the reader's nose. Storebought posters and packaging are commonly printed at relatively course halftone rulings of 100 or 120 LPI, in which cases 180 PPI is a gracious plenty (especially for graphic artwork); resolutions as low as 50 PPI are commonly used for billboards.

With the advent of finer-resolution mobile device screens, the graphics industry is finally beginning to wake up to the above-stated principle by referring to resolution in terms of PPD (Pixels Per Degree), which at last takes viewing distance into consideration along with image size. PPD refers to a degree of the angle of view which the image will occupy. The human eye can only resolve so many discrete colors of a given size at a given distance--in other words, we can only see so many pixels within a degree of our field of vision.

Have you noticed how much of this whole issue of raster resolution boils down to a basis of common sense? Adoption of and familiary with the principle of PPD could relieve much of the common resolution confusion. A billboard you see when zooming down the road on your Honda Grom is intended to be readable when it occupies roughly the same angle of view as does a printed page held in your hand. PPD is a resolution equalizer.

JET

[pauland]

Mike, thanks for the images. Out of curiosity I just took the smaller image and enlarged it using the OS to compare.

I think we can both agree that (as we might expect) edges and crispness in general are degraded by the resize and we can see signs of jpg artifacts in the zoomed image.

I can say for certain that the image that you zoomed Mike, was better than mine was - better edges.

I can see why you'd use this kind of software - both you and Acorn - but as you say I think this is a case of doing it when there's no alternative. There's no doubt that quality is lost and it's really sharpness that we lose down to the resampling. I can see for myself that the edges are enhanced by the specialist software to try and keep as much definition as possible.

Interesting.

I remember that there's a company that uses image analysis and fractals to get some impressive results, but I forget which one now. I know it used to be expensive.

[csehz]

JET thanks for this detailed explanation

[mwenz]

...I can see why you'd use this kind of software - both you and Acorn - but as you say I think this is a case of doing it when there's no alternative...

I rarely use Smilla to go larger than 200%. Most of the time I use it is when a placed image in a page layout application has to be sized up and its effective DPI drops below 150, the job is for print and there is no better image available. Then I'll use it for an increase of 125% to 200% just to get the effective res higher. And I'll use the lowest percentage to pump up the effective res that I can get away with. In those circumstances, it is quite effective.

In all cases, the first thing to do is to obtain an image with a greater PPI if at all possible and proceed from there.

I remember that there's a company that uses image analysis and fractals to get some impressive results, but I forget which one now. I know it used to be expensive.

There is a PhotoShop plug-in that I used when I worked for another company years ago. It was the best at this. It was fairly expensive and I don't remember its name. I have also used PerfectResize from onOne software. It does a good job on most images, costs about $150 for a new license. On some images it is better than the free Smilla, but on most I tried it is only marginally better when resizing at the percentages I will use. It does, however, go to larger increases far, far better. The newer versions also operate as a stand-alone application so one does not have to have PhotoShop.

Mike