Jagged Edges

[dustin_smith1]

<p>Anyone know how to both avoid, and/or correct having the edges of objects in photos come out jagged and pixelated? I have posted a pic of what im talking about. The edges of the window on the crane, as well as the cable hanging down is very jagged and distorted. Thanks.</p>

[dustin_smith1]

<p>pic</p>

[kou_lee]

<p>They look fine to me ... If your camera uses digital zoom, that will cause what you are describing.</p>

[jimmckinnon]

<p>Dustin, I have to agree with Kou Lee, the image doesn't look that bad in terms of edges and distortion to me. I'm not seeing a problem with the window edges, and what we're seeing on the cable is probably just a result of the strong back-light - the bright sky. </p>

[tangibleimages]

<p>One key thing to keep in mind is that you are dealing with digital photography and its square pixels. Any diagonal line, especially those of high contrast, will render as a series of squares that touch at the corner.</p>

[protest]

<p>Describe your workflow, i.e. what did you start with, and what steps did you take in processing the image?</p>

[kevin_delson]

<p>While there is some CA, I see no jaged edge or pixelation.</p>

<p>Make sure you are viewing your images at 100%..NOT 120%..NOT 45% etc...</p>

<p> </p>

[matthew_rusbarsky]

<p>On my monitor at high res, I can see the jagged edges. I'm guessing an artifact of JPEG compression. View it in TIFF or try another JPEG algorithm?</p>

[Ray House]

<p>Dustin, zoom in on the jaggies and they will disappear.</p>

[lex_jenkins]

<p>You're seeing a combination of factors that degrade the image:</p>

<ul>

<li>Primarily, artifacts from compressing a full resolution 15 mp file to 250 kb.</li>

<li>Chromatic aberration in high contrast areas.</li>

</ul>

<p>I'm attaching a crop magnified to illustrate the artifacts and CA.</p>

<p>First, don't compress a 4,752 x 3,168, 15 mp JPEG to only 250 kb. A high quality JPEG of that resolution would be closer to 2 MB or larger.</p>

<p>If you need to reduce a JPEG to a specific file size of 250 kb or so, you also need to reduce the dimensions. Otherwise, given the current compression technology, artifacts are practically unavoidable.</p>

<p>Second, CA is hard to avoid in high contrast areas. Try stopping down to f/8 or so. At f/4 you may be close to maximum aperture for your lens, which tends to show more CA (oversimplifying here for the sake of brevity).</p><div></div>

[tom_mann1]

<p>Dustin -</p>

<p>I'll bet that the reason you are seeing "jaggies" is exactly what was suggested above, namely, you are almost certainly not viewing the image at 100% magnification (ie, one camera pixel = 1 screen pixel). </p>

<p>When you are viewing a photo at typical lower magnifications, e.g., "fit to window" or browser default magnification, whatever software that is being used to display the image has to figure out how to map the value of several pixels from the original into the value of a single screen pixel. There are many, many ways to do this, and figuring out which is the best of these is a huge subject of discussion. Some of these methods (eg, "nearest neighbor") are extremely fast, but always produce staircase edges on diagonal lines. Other methods (eg, "bilinear") are slower, and produce smooth diagonal lines which, unfortunately, are much too soft. Still other methods (eg, "Lanczos") try to find a nice compromise between the above two extremes. </p>

<p>The ONLY way to eliminate the uncertainty and artifacts introduced by your image display software is to view use "100% magnification" when evaluating each image. This completely removes the need for the above software and takes it out of the loop.</p>

<p>Once you are sure the above issue is not making you believe that there are jaggies when there actually are none, then you can get into the issues that Lex brought up, namely, JPG compression artifacts (ie, block-like artifacts), the various types of chromatic aberration (ie, colors on high contrast edges, some of which get worse as you go out from the center of the frame), etc.</p>

<p>HTH,</p>

<p>Tom M</p>

<p>PS - I don't see objectionable staircase-like "jaggies" on your image, only the problems that Lex mentioned. However, if I try to display your image with my browser, using its default "fit image to window" option, I sure as heck see "jaggies". The problem is that they are caused by my browser, not intrinsic to your image.</p>

 

[James G. Dainis]

Tom is right. I see the jaggies when the photo is at "fit to window" but when I click on it to get it full size, the jaggies are gone. Full size is about three screens wide on my computer. I can only see about 1/9th of it at a time as I scroll back and forth and up and done. For web display you should only have the image about 700 pixels wide for easy and full viewing. Resizing it done before posting it would avoid the jaggies.

[scott_murphy_photography]

<p>"Stairsteps" are a necessary evil when it comes to digital photography. It stems from the fact that pixels are square instead of round like film grain.</p>

<p>The best way to minimize them is to save the file with no compression and limit the amount of sharpening you do to an image, as that can really exacerbate it.</p>

[tom_mann1]

<p>[James D]: <em>"...Resizing it done before posting it would avoid the jaggies..."</em><br>

Absolutely, provided, of course, you use an appropriate resampling algorithm, ie, something like Lanczos.</p>

<p>[scott M]: <em>"... "Stairsteps" are a necessary evil when it comes to digital photography. ..."</em><br>

Primarily when you or an algorithm that you are employing (often unknowingly) is trying to eke out the last bit of resolution from an image file. </p>

<p>[scott M]: <em>"... It stems from the fact that pixels are square instead of round like film grain ..."</em><br>

Not quite. If you generated digital images using round pixels instead of square pixels, you would still have exactly the same problems with stair-stepping. The only difference would be that the "stairs" would be differently shaped, ie, more like the outline of a cookie-cutter instead of stairs. More to the point, pixels in a digital capture are all exactly the same size, are spaced in a perfectly regular array, and can now be easily magnified to the point where their sharp boundaries are visible. </p>

<p>Not to bring up the classic film-vs-digital arguments, but this is to be contrasted with film in which, the grain particles are not all the same size, are not all oriented perfectly, are not even the same shape, are spaced randomly, partial overlap of grains at different depths, and because they require a decent microscope to be seen instead of a single mouse click. There has been a huge amount written on this subject. For example, if you have the stamina, you may want to wade through this thread from Nov. 2008:<br>

http://www.photo.net/film-and-processing-forum/00ROOo</p>

<p>Cheers,</p>

<p>Tom M.</p>

<p><em><br /></em></p>

[lex_jenkins]

<p>I hadn't looked at the browser-resized version of Dustin's photo but, yup, he may be seeing scaling artifacts or aliasing as well. Very common problem with browsers and image viewers when changing magnification to view the entire image onscreen.</p>

<p>I'm attaching a comparison between the scaled-down version I saw in my browser and a resampled version of the same dimensions (all resized to fit photo.net's 700 pixel width limit). There's very definite stair-stepping scaling artifacts in the screen-capture version. It's still noticeable in the resampled version but not quite as severe. You can probably get better resampling results from a higher quality version of the original. But some stair-stepping aliasing and jaggies are unavoidable when images are resized to small JPEGs for web viewing. It will be more noticeable in thin lines, ropes, etc., or hard edged diagonal lines.</p><div></div>

[tom_mann1]

<p>Lex, don't forget that when a browser scales an image, it must also perform some sort of resampling. The question is do you want the resampling to be under your control or under the control of the people who designed the browser. The goal of the browser folks is slanted more toward execution speed; photographers usually want higher quality images. The two goals usually suggest algorithms on opposite sides of the quality-speed spectrum.</p>

<p>Tom M</p>

[lex_jenkins]

<p>Yup, tho' I don't know whether other browsers handle this more gracefully than my old version of Firefox 2.x. Zorro de Fuego Dos Equis tends to crush-fit oversized images to fit the browser window, with much more obvious scaling artifacts than user-chosen resampling. I see it pretty often when folks embed their oversized images hosted off-site. Pretty ugly results when a large image is crush-fit to suit photo.net's 700 pixel width limit.</p>

[tom_mann1]

<p>Lex - I'm using Firefox 3.6.3 (...the latest version, I think) and I also see plenty of artifacts when it down-samples an image to fit in a window. To adapt the old saying, <em>"you can have fast or you can have good -- pick one"</em>.</p>

<p>BTW, I don't think that all the variation in width and darkness of the LH vertical line (ie, the line that your upper arrow points to) is due to resampling artifacts. If you zoom in on that line in the original full-rez image, you can see that it is actually a twisted line with periodic bumps and a wider section near the bottom where the individual strands that make up the line splay out a bit. After down-rezing, I think we're seeing the remnants of the true structure of the line.</p>

<p>For yucks, I tried my hand at manually dow-rez'ing the image. I first used Topaz de-JPG to try to get rid of some of the JPG artifacts. Then I used simple bicubic to down-rez to 698 pixels vertical, followed by my usual "sharpen for web" routine and a pass with Neat Image to reduce the banding in the sky. The result is attached below. It's not much better than yours.</p>

<p>Cheers,</p>

<p>Tom M.</p>

<p>PPS - I had not idea Firefox really is translated as "Zorro de Fuego". I assumed that being a proper noun and probably a registered trademark, "Firefox" would stay as "Firefox" even in Spanish, and you were just having some fun with the cute mental image it evoked, ie, a little "Zorro" masked fox with a streak of fire behind it and a Dos Equis in his hand. :-)</p>

<p>PS - The OP hasn't posted again on the thread. I hope we didn't drive him off with all this minute examination of the back of our fingernails. ;-)</p><div></div>

[lex_jenkins]

<p>Good point, regarding removing the JPEG artifacts before resampling. Presumably the resampling algorithm won't know the difference between valid data and artifacts, and might have created more noticeable aliasing in my quickie attempt.</p>