In a discussion about jpeg artifacts and their effects on the quility of images posted here, I'm uploading 3 images all of different settings, but I'm not telling you what those settings are... Please examine them carefully and tell me which you think is the better QUALITY.. Thanks for your time. The Bryster

Hummmm, It's hard to judge while trying to scroll from top to bottom but I think sample 2 shows more latitude.

All 3 look exactly the same to me.

I'm with Aldaron, they all look the same to me also.

guess I need to have my eyes checked..could you bring them all into one pic, maybe? I dunno...also, if there's something different going on in each (types of jpeg, etc), that might give a hint as to what we're looking for. They all look good to me, Btw..

Not enough mid tones for me to tell.

Well the thing is they all look great. it's very easy to get a great image with a good jpg file size. So forgive my ignorance in that why this should be an issue. LB

I saved them as BMP's and differenced them. The only place I could really see a difference this way was in the subscription. The differences are so minimal that I doubt you can see them just by looking at the pics.

I'm with Pakled...." could you bring them all into one pic, maybe? " Hard to see a difference when scrolling through them.... The are good looking pics though.

As Aldaron said.

I agree they all look the same. One thing though, is that with the sun where it is in the picture, shadows should be towards the viewer. Did you omit the shadows on purpose?

Well, they all show being saved at a .jpg quality of 99% or 100%. Perhaps the difference in file size came with various sub-samplings, progressive, removing the exif info, or supressing the jfif header, etc. Sample 2 comes up with 32,276 unique colors, the other two samples both come up with 31,151 unique colors. Other than that, even viewing all three at once under 8x magnification, I can tell no true visual difference. AgentSmith

*Note; Actually, Sample #1 and Sample #3, come in at the same exact file size (161,518), and the same exact unique color count, so technically, they are the same exact .jpg pixel per pixel. AgentSmith

Usually JPG artifacts show in the compression level, not necessarily the compression type. (Aren't JPGs all some LZ7 derivative anyways?) There could be differences in embedded information such as watermarking, but most of the time that isn't visible. Usually JPGs will show their flaws in high contrast areas (like that around text.) The differences in your pics here are too subtle to really notice.

WOW! So many replies! First off: Thank you all very much for the time you have taken and all the work you have done in viewing these images. It really is appreciated. Thank you so much. If we look at where the hills meet the sky, the artifacting seem most obvious. The trouble is, the pics are supposed to be at different resolutions...#2 is supposed to be at 300dpi. #1 & #3 are supposed to be at 72dpi. There is, as you all seem to have said, little or no difference in the quality of the images as seen here on the Rend. So my question now is: What is going on? No matter what I do, I can't seem to find a formlua by which I can guarentee that my pics, although looking good here at home, will look as good on the forum. In a previous thread it was suggested that I change the dpi to 96 instead of the default 72, or, render to disc, or change the print rez....... and I'm not the only one with this problem. I've seen loads of pics here with artifacting, some worse than mine, and that's a shame. I always save my pics in PSP7 with the Jpg set at 'Best Quality' but nothing seems to work. Perhaps someone has the solution already. Who, out of all of us here, always gets good results after posting on the Rend? Perhaps you could tells us how you do it? Thanks once again.... The Bryster

All 3 of your pics came up 72 dpi for me. In IrfanView, Thumbsplus and Photoshop. They all say 72 dpi. From what I have found and seen, rendering a pic in 72dpi vs. 300 dpi, makes absolutely no difference whatsoever. There is no set formula for jpg's. I would highly suggest using something to save your jpg's where you have a numerical setting for the jpg quality, not just "Best quality". That's too vague, imo. But, to be general...overall...save at 90% quality. that is usually the overall sweetspot. If artifacting shows, start taking it up to 95%. RARELY will you ever see a difference between a pic saved at 95% Vs. 100%. It takes a lot of straining of my eyes, to see any difference between my own .bmp pics on my hard drive Vs. the jpg versions on Renderosity. Sure, zoom in on contrasty spots, but at looking at it, at 100% zoom, with the naked eye...no difference. AS

OK, to further explain the technique, let me offer up the source. It comes from a tutorial in 3D Commune so I cannot replicate the source in total here because of copyright issues, so for those who wish to see the tutorial in it's entirety, you must first be a registerted member of that site; Http://www.3dcommune.com/ . It's a great site and I highly recommend it. The tutorial itself is called "Sky Effects For All" by Paul Pappathan a/k/a Pinhead could be found here; http://www.pinhead.robbes.com/tutorial/Skyfx/sfx3.html The technique is explained down around the end of the tutorial in a section called "Time to render". He explains; "Do not Render as jpg...they suck. Save all original renders as PSD, BMP or PICT then convert to jpeg for posting. Your results will be more controllable and look 10 times better. Really and truly. ...Here's How to do it right! ...The idea is to render large to make smaller. So make it big! Your images will look much, much better when resampled as Jpeg. ...if you're going to post an 800 x 600 image, your render settings should be: 1600 pix wide x 1200 pix high x 96 dpi minimum. 96 dpi is now the effective resolution of consumer monitors. Not 72". This makes total sense to me and is in agreement with what most other graphics authorities say on the matter. I've tried this technique and it works for me; http://cg-quest.com/End_table.htm

True. As Paul says in his tutorial "if you want high quality PRINTS...this is the way to go". Prints. But, if you're just rendering to upload to your gallery, render the exact size you are going to upload. Otherwise, your pixels will look resized. And, to render any size scene twice as big...will take 4 times longer in render times. But, if you are going to print your pics to start with, always render in larger dimensions. AS

I will admit, most people who are not me...render with printing in mind. I for some reason, have never printed anything of my own, so I am always rendering with just my gallery in mind. I figure if I ever need to I'll render big (and for MANY hours) later on. What I need is a job that has great printers that I can ahem..."utilize". ;o) AS

hmm..maybe that's why I don't get the sharpness I want..I've reduced some pics down lately, just to get around lengthy renders..

I'm sorry, but I beg to differ. We all know that to publish to print we need to up the dpi to somewhere between 150-300 dpi (or above, this is a standard rule of thumb), that goes without saying. Unless I'm reading the tutorial wrong, he specifically says; "if you're going to post an 800 x 600 image, your render settings should be: 1600 pix wide x 1200 pix high x 96 dpi minimum. 96 dpi is now the effective resolution of consumer monitors. Not 72." He's referring to monitor resolution. 72dpi had always been the standard for web image viewing (as per HTML standards), correct me if I'm wrong. LOL...these discussions on image resolution always get so entangled and confusing.

Sorry again AS, our posts crossed. I now see your illustration but the question I need to ask is, was the 400x400 render at 72 dpi or 96 dpi?

Funny thing, AS, that with jagged lines happened to me with "Columbia Ad Astra". In Bryce. You can't see it, because I twiddled it in Photoshop, but the lower edge of the wing was jagged just like the line in your picture. And now I'm resizing it by half and the line is smooth. The same thing happens if I resize by 25% or by 12.5%. Anyway, Bryster, what apprently happened, PSP didn't set the resolution of the second picture to 300 dpi. It is slightly bigger in file size, but very little. All three of them are at quality setting of 8, when Photoshop can go to 12. And if I wanted to resave them at that quality, they would be at about 53-54k in size. So, it's no wonder they all look the same if they are saved at the same quality setting. I also zoomed in at 1000% and couldn't see any banding or artifacts. I wouldn't expect any at 153 kB. :-) There is a nice little program called Image Optimizer at http://www.xatio.com/ You can do wonders with it for jpgs and gifs and I was using it before I got Photoshop 6. IO is possibly still better, but the Save For Web feature satisfies all my needs now.

Looks like I've entered a minefield here! From all these differing opinions it seems as if I have to experiment to find what will work for me. I had a similar problem when trying to save a Crossword Puzzle grid and transfer it to a page on MS Publisher. I had to mess around for ages to get the grid numbers to look anyway near decent. OK, so rule one seems to be: if you want to print to paper, render at twice the size........ and rule #2 is............??? Sorry! I'm too confused now. I'll sort it out in the morning. I have sciatica and I'm out of my head on co-codamol...LOL Thanks for all your suggestions, Guys! The Bryster

sorry, the URL for Xatio has been expired, relocated, or something, just popup after popup..que lastima. Let's try a different approach..;)

Attached Link: http://www.xat.com/internet_technology/download.html

Note; They both have the jaggies but sample B a lot less so. I've run this test several times, repositioning the model everytime with the same results. To the scene I added one radial light set at 35, white, shadow ambience set to 100% black.

Oh, absloutely! IF your render already HAS jagged edges, reducing the pic by half will help somewhat. Some use this as a very rough way to get around having to use the last rendering pass, the AA pass. It doesn't compare to having a real AA render, but, it's good for getting a rough idea of what the final scene would look like. BUT, if you have rendered normally...(with your scene anti-aliased), and you then resize your pic by half, then you will get small sharp/jagged pixels as a result. (almost always) And, since rendering with AA or without AA, and then halving, has always made pics look bad COMPARED to a pic rendered with AA and NOT resized...I would suggest not resizing pics, just to get the clearest picture. All just imo. Dpi doesn't even enter the "picture" so to speak for me. The exmaple I gave above was at 72dpi. But, I could give you the same example at 72, 96, 150, or 300 (if you want) and there will be no difference in file size, color count, render time, or visually. I could do more comparisons, but this is what I have experimented with and found out myself. Just a difference of opinion, but I have yet to find a render look any different when made at different dpi's. (on screen) Quest, you could do your sample again with each pic having any dpi you want, it's going to turn out the same. So far, it has with me and my little experiments. But, everyone has different pc's, different monitors... Paul does say at that tutorial "Make Output Size at least 2X of what you want to post for the web". He never says to RESIZE, he says "Your images will look much, much better when resampled as Jpegs". RESAMPLED, not resized, big difference. Unless I'm wrong, he means when you resample (convert) your bmp to a jpg. And, he's right in that aspect also. Large dimension bmp's can have a smaller jpg quality and still look good (on screen), compared to a smaller bmp saved at the same jpg quality. I'm not sure about dpi and computer monitors, I've never seen dpi listed as a spec on any monitor. They use dot pitch as their level of measure. But, I have seen that Paul is a WELL informed dude, I wouldn't doubt what he knows and what I haven't heard of, lol. But, you're right Bryster, with .jpg's ALWAYS experiment, always, lol. AS

*Footnote - In fact I have a Photoshop formula somewhere for taking an UN-anti-aliased render and using the resize and various filters to make it look fairly close to a regularly AA pic. I'll have to look for that. AS

AS, just FYI....resample and resize are the same thing :) Not sure why it's called resample but it is in Photopaint.

Er, Aldaron, it's not. Photoshop help: "Resampling refers to changing the pixel dimensions (and therefore display size) of an image. When you downsample (or decrease the number of pixels), information is deleted from the image. When you resample up (or increase the number of pixels), new pixels are added based on color values of existing pixels. You specify an interpolation method to determine how pixels are added or deleted." The problem is especially bad when you increase the size, cause PS will add information which was not there. But if you turn the resampling off, and then change the resolution, the pixel dimensions will stay the same, but the metric (inchic :-)) dimensions will change. That way you can see how big your image will be in print, for instance. You can also change the physical dimensions, and the pixel dimensions will stay the same. You'll see the same picture on screen, but what's printed is smaller or bigger.

Ah thanks Erlik. I never really knew that's what it was doing. In Photopaint it lets you load an image by resampling but you can't go above 100% so I just thought it was to resize it smaller. Nice to know it's actually taking away pixels, I'll have to remember that.

I render all my work at 1280x1024 and save as bmp. I then post work in PhotoPaint and resave as LZW tiff (for my wallpaper use). I then resample (w/anti-aliaing) to 1024x819 and crop to 1024x768. I always apply a 50 to 85% unsharp filter to compensate for any induced blur. I export as a jpeg at about 15-20% (inv proportional to PS) with 4:4 sampling and optimum settings. Not sure what the Optimum means but I do see the difference in image qual between it and regular. I always adjust the compression valueb checking the on the fly preview/original viewer. Aldaron- In PP if you check the "keep image the same" size box in the resample popup all you do is alter the dpi ratio in the image file header with no adittion or deletion of data. If you leave this box unchecked, you will add or remove pixel data to alter the image size.

Incarnadine: I understand just about nothing of what you said...........LOL Complicated or what?...... Thanks, Mate (No disrespect)

None taken! I do my postwork in PhotoPaint which does not behave the same as PhotoShop. All my posted images are compressed to the equivalent (theoretically) of 80 to 85 in PhotoShop. PhotoPaint works on a data point by data point basis where a data point equals a pixel at 100% size. The actual output scale dots per inch value is contained in the image header data but does not usually enter into any processing/editing work which is all based on the data point info.

i.e. a 1024x768 image is an array of 1024x768 data points whether at 300 dpi or 72. Only the output inch size varies, not the data array.

Er, no. (for the second time in this thread. :-)) Dot and pixel are not one and the same. A pixel can be made of several dots of different colours, while a dot is always only one colour. So, if you have 300 dots per inch, you have roughly four times the data you have with 72 dots per inch. And that's the reason why simply increasing the resolution* from 72dpi to 300dpi in an image processing program ends badly. The program interpolates by its own calculations and puts in what you didn't intend to be there. * (rant) That is resolution. 1024x768 are dimensions. (rant)

Lol, I think we need a jpg forum. (it's almost a science in itself) AS

Aaaargh! rendo threw away my reply! and I was even polite!

I agree, 1024x768 are indeed dimensions. They are not print dimensions or pixel dimensions though! Corel PhotoPaint (and almost all other image manip tools) handle an image as a matrix of data values of y rows of x data points of 24 bit rgb cluster values. Additionally at the beginning of the image file is a header with the encoding algorythm parameters and the output scale factor info. This output scale factor is what actually sets the print resolution. Display resolution on the other hand is based upon the maximum density of the rgb phosphor elements on your screen (as determined by the phyical limits of putting all the holes in the shadow mask.)That density determines the ultimate resolution where each cluster equals one visual display pixel. Now if you have your display set at a resolution below that maximum, you will have each visual pixel (representing one data point in the array) comprising several rgb clusters. In Corel PhotoPaint there is an option under Resample to alter only the output scale factor and not the matrix of image data. If this is uncheck then the resample command acts directly on the data matrix and either interpolates down (removing data points, causing some degree of blur) or interpolates up (adding data points, causing some degree of pixellation). Jpeg compression works by summarizing a string of identical valued data points as the number of consecutive points of precisely RGB value x and throwing away the actual matrix values. Increasing the compression factor very roughly translates into toleranceing that value of x. This starts to corrupt the purity of the colour accuracy and with the loss of the data cannot be recovered from. It is these corruptions that we refer to as artifacts. Truce? (grin)

Hmmm... thinks for a moment... I thought the DPI settings were for controlling print quality. More DPI means a sharper crisper printout (provided the printer is any good.) The DPI settings in most 2D graphics program more or less handle how such high resolution images are viewed by the program user. It basically compensates for the image size so that what appears to be about an inch on the screen is an inch on the printout. So that way a 300DPI image could be viewed on the screen. For most purposes though, a dot is roughly equivalent to a pixel. If that 300DPI had the view setting switched to 72DPI in the graphics program (or was viewed using a program that didn't use the interpolation data) you'd more than likely get a huge image that would require some scolling. Notice that the DPI value doesn't really change the amount of dots (or pixels for that matter), it just tells the computer how to go about displaying it or printing it. Either way, with a high or low DPI setting (independant of initial creation, as that sets the pixel or dot density provided initial dimensions are entered in in or cm.) - I'm willing to bet that the file would be the same size. As far as JPGs are concerned when it comes to bigger images - the quality loss becomes less noticable. This is because its data loss occurs at the dot or pixel level. The bigger the image density, the closer you have to zoom in to see the artifacts. It will always form 8x8 pixel (or dot for that matter) blocks once the distortion becomes evident. Dunno if that helped any, but that's my 2.

Yes generally DPI is to control print quality. We must be careful to distinguish between Dots Per Inch (printer resolution), Display Pixels Per Inch (screen resolution based upon video card driver settings; based upon the defined number of pixels divided by the width of the display) and the absolute display resolution (a constant value, fixed by the hardware of the display; typically .22 - .25 mm between RGB clusters, therefore approx 96 PPI maximum). A 1 inch 300 dpi image will display as three inches of display width at the 100% size setting. A 1 inch 50 dpi image will display as 1/2 inch of display width at the 100% size setting. In order to display a 1 inch 300 dpi image as 1 inch of display width you (or the program) would have to create a temporary (so as not to damage the image data) down interpolation to the 96 PPI hardware limit. This is why "resize to fit" often introduces artifacts similar to those of the jpeg process.

Erlik- I have reread your previous posts and I think we are actually in agreement, just stuck on different terminology. I use Corel PhotoPaint and I expect you use Adobe PhotoShop? Thus my selectable function resample is the same as your resize or resample!

Yup, Photoshop. And yes, we are in agreement. Truce. :-) Just one more thing: Paul, I don't understand what you mean with "it will be the same file size"? That the file size will be same whether you have a 72dpi or 300 dpi image? In that case, it certainly won't.