Attached Link: Original Bench Picture

I am new at this and just started uploading to RR. In your answer try to dumb down the tech speak, you old camera guys and gals talk like rocket scientist...FOV...POV...Fstop here..aperture there...you get the drift....LOL :) ..... Go easy on me, if you can.   I have questions, I got a lot of comments on a picture I uploaded, "A Cool Bench to Sit On", was out of focus, and when I went back and looked at it online it is definitly blurry. Then I thought about it and did not remember it that way in the original, before I cropped and resized it. So I went and looked, the original is clear and crisp. The software I use to crop and resize came with my laptop. (HP Photosmart Premier) The original picture size is 2592x3872 @4.64mb and I had to resize it to 670x899@479kb. I have to keep resizing until I get to 512kb or under to upload. Here are my questions, when you resize a picture do you lose clarity?  Is there a better way to resize and crop  and not lose clarity. What do you think? **I appriciate you taking time to answer, and want to thank you in advance.** I also want to thank Richardphotos and  auntietk  who have already given me advice to try on this subject.

Todd

When you resize you do lose clarity, most software has different ways of doing it so low / medium / high quality settings (you may want to take a look through the software helpfile).
 
Also when you save to a new file you may add blur due to the compression of the data (the way the data is made smaller for the final file). Certain formats such as jpg use a compression that will blur the image more every time you save it and it has quality settings too (a trade of quality versus size), so while working it's best to use an image type that is lossless such as PSD(photoshop), PNG or XCF (there are loads of others) and only convert once to jpg

You can look atit this way: in the digital world, an image is chopped up into pieces (like a grid, a chess-board). Each of thiese pieces, called "pixels" has a certain place and a certain value of lightness for the three basic colours Red, Green and Blue. When those pixels are small enough, we interpret the grid as an image (just like on TV).

Now what happens when you have a grid of 2592x3872 of these pixels that has to be reduced to one of 670x899 pixels? Clearly, the software has to try and get rid of more that three out of four pixels, but in such a way that we still interpret the result as the same image. The way it does this is by searching for pixels that resemble one-another and substitute them with a pixel of a kind of mean hue and lightness. But!!! Software does not "see" the image: it's only math that is calculated. Result: it is not only always simpler (less pixels in the grid) but also less shrp where there were crisp transitions. So it looks to be more blurry.

So what must be done is try and get, once again, crisper transitions. This can be achieved in two ways. First of all to reduce the size in several smaller steps. Then, for example, the software does not have to calculate one out of four, but, say, two out of three pixels. MUCH better to do, and easier to calculate. You do this by choosing intermediate values like (only an example) 2592 to 1913 to 1436 to 1121 to 899.
A second method is by so-called "sharpening". I can't tell anything about HP software, but I do can tell you that if you're on a Win computer, you can download Irfanview, a free image viewer with very (!!) good mathematics under the hood. What's more: it's free and very easy to use. And it accepts Photoshop plugins, many of them free.

Irfanview is also a master in saving as jpg, the main format for photographs on the Internet. Jpg also reduces the file-size, but that's another story. It's best to save your pics as tiff for keepers, and set them to jpg for the web. All this can be done in the free Irfanview.

One last thing: Increase or decrease lightness, give more or less saturation, crop, resize,...anything, but keep sharpening untill last.

Feel free to ask any more questions. There's no stupid question, only an incomplete answer.

Tanchelyn pretty much covering things, with the help of Gog. Let us know how it goes -- good luck!

Yeah Tan has covered pretty much everything.

Personally I would suggest cropping/resizing in stages and at each stage do some sharpening, but don't go over board.

And get yourself some decent software, the standard HP stuff is shocking! Maybe not PhotoShop but Arcsoft's stuff is pretty good. Elements is cheap £50 I think...

BTW - I see that your shot is okay for focus from the link you provided...

I would strongly disagree with PeeWee05. :ohmy: :biggrin: Do as few resizings as possible and preferably only the one to the final size. If I have to produce several sizes, I reload the original large file each time and start again from there.

http://www.youtube.com/watch?v=vIFCV2spKtg

Found this over in the Bryce Forum.
Take a look.

Yep, the plugin used in that vid is linked in my thread here in the photo forum...

http://www.renderosity.com/mod/forumpro/showthread.php?thread_id=2714299

It'sa great toy but I haven't used it in anger yet...

quote: "Do as few resizings as possible and preferably only the one to the final size."

Reducing size in steps is the only way to obtain a good result. You'll always lose information / image quality but less when the average that has to be calculated is 1/ not a simple division (like one out of two) and 2/ the number of pixels that has to be eliminated and replaced is as small as possible.
You have of course the right to disagree, but please give us a sound foundation. This is a non-tech user and I find your entry very confusing.
So please explain as your opinion goes against all mathematical logic.

I can see both sides of the arguments and it comes down to the complexitys of the maths,

Without getting too deep in the math, the higher quality resize algorythms use cubic or bicubic formulaes, the software analyses each colour and tries to plot a curved line that gives a peak at where the chosen average colour is, problem is there will basically be errors, the more resizes the more errors in the maths. On the other hand when a linear algorythm is used (often the lower quality option in software), it means straight lines rather then curved apporximations, so in some circumstance lots of small steps will approximate closer to a curve and get you closer to the right point.

Hi Tanchelyn As Gog says, its the complexity of the resampling algorithm. Though the algorithms vary widely from program to program (even when they use the same name, 'Bicubic'!) some of them use more than 1024 adjacent pixels for sampling. So if you've taken away some of the original data with an intermediate resize and then gone and altered it further with some sharpening etc. You're actually reducing the efficiency of the algorithm. (I just noticed, sharpening before reduction!!! :scared::scared::scared: The accepted wisdom is to do a slight blur before reduction to avoid moire.)

I can agree on moiré. When you have a finely detailed texture with a repearing pattern that comes close to the resolution, you can get moiré effects. A workaround is indeed a slight blur. It's not a    soltution as you lose detail, but when done by someone with experience the slight disadvantage outweights the moiré pattern problem.
Bicubic uses 4x4 pixels: horizontal, vertical and diagonal. It's considered to be the best balance between calculating time and result. There are better algorythms, but these are often linked to expensive plugins or standalones.
But the problem remains the same: making the grid smaller means averaging hues. Of course it's better to have each pixel "watch round itself" to every pixel that surrounds it (bicubic) but if this happens in long series -to get rid of a lot of pixels in one step- more image is lost that when you do the same in several smaller steps. True, you may lose more in the rounding to integers, but you win more than you lose.
Look at the image: the left pattern was downsampled to half in one step, the right one in four steps. If you look closely you'll see how the four steps create a nicer set of pixels to come close to the original.
Hey, I didn't invent this! Just read what most Photoshop gurus write. It's from them I learned this trick. If you have to resize only a little, this doesn't matter. But if you resize a lot it's visually better.

You always lose detail (information, data) but less when you take the stairs instead of jumping through the window.

Here's the image.