Optimum Uploading size for images.

Hi Rubber

I sell mainly postcards and like to show them as large as possible so that my bidders can see exactly what they are bidding on in regard to the subject and the card's condition.
To obtain this size image from a postcard of 5.5 inches by 3 inches I scan in at c.200dpi.
This gives me an image of around 1093 x 719 and 350kb approx.
Using auctivas uploader which optimises [and compresses?] this image to 1024 often increases the file size!
If I scan at 200dpi and then save at 72dpi to get an image file of around 150kb and then use auctivas uploader the file size still increase a tadge.
I have also noticed that the auctiva uploader appears to have increased it's compression rate as my images all now have that low res sort of blurring which is not what you would expect with a higher file size!
It'a mystery as to what is the recommended way of getting your images to display at their best on Auctiva!

Even images that I self host into auctiva do not display as well as they do on my hosting site!

Hi,

Here's a little contribution to the discussion....

I decided to run a simple test. I took a photo fresh from my Camera. The photo was done at the 6.1MP setting with Standard jpeg compression on my Kodak Easyshare Z760. Note, my camera also has a Fine setting which decreases compression (larger file size for same shot). The Auctiva/Aurigma uploaded the photo and it was stored at 167.42KB.

Here's the photo at 400x300 px as stored at Auctiva.



I then displayed the photo in a browser window by copy and paste of the link, and saved the photo to my desktop.

Here are some stats on before and after, after being the saved photo back to my desktop.

Before: Size 1,257,942 bytes, WxH 2856x2142 px, Res 96x96 dots per inch

After: Size 171439 bytes, WxH 1024x768 px, Res. 230x230 dots per inch

It was the dots per inch that caught my attention. I checked the photo before and after with MS Paint, which does give a DPI attribute for Properties.

That increase in DPI might account for the dithering effects seen on photos and the size increase.

When I get a chance, I'll try a similar test with a scanned image at high scan res (high DPI).

The Fine setting on my camera does not affect the DPI, only the compression (tested).

Danno

Edit: I took the same original photo and resized it using Paint to 1019x767 (just under the 1024x768 limit) and used the Basic HTML uploader for a test. Here's the results.

Before: Size 131,700 bytes, WxH 1019x765 px, Res 96x96 dots per inch

After: Size 124,097 bytes, WxH 1019x765 px, Res. 96x96 dots per inch

Note the DPI did not change and only a small decrease in file size was observed.

Here's that second test with pre-sized photo at Auctiva 400x300.



The second test suggests that presizing a photo that requires best resolution (precludes dithering) and using the old HTML uploader is possibly a better choice when extremely fine details are required, i.e. it appears to upload and store as processed at the desktop.

Sophie's photo when displayed at the full size, as stored by Aurigma, might be a whisker better (pun intended).

Edit2: I examined the original photo and the two saved photos from the prior testing with Picasa.

Picasa reports that the original photo has a "JPEG Quality" of 94(411). The photo resized by the Aurigma uploader shows 90(411). The photo resized by Paint and uploaded shows 75(411) for both the before and after. The Picasa documentation is a little vague, but I'm interpreting the quality as a measure of compression (as percent of original retained). As I suggested, it appears that the Aurigma reduced photo might be a whisker better. However, you do get the more exact copy by using the HTML uploader.

Edit3: One more test and a surprise.

I decided to use Picasa to resize the photo to exactly 1024x768 (same as Aurigma after upload) and maintain the quality. The results are as follows:

Before: Size 262,005 bytes, WxH 1024x768 px, Res 96x96 dots per inch, Quality 94(411)

After: Size 115,173 bytes, WxH 1024x768 px, Res. 96x96 dots per inch, Quality 75(411)

Note the major decrease in stored bytes and the drop in quality from 94 to 75. It appears the HTML uploader further compresses the image after upload, so it would appear the Aurigma uploader maintains a higher quality.

Edit4: OK....I had to give Picasa - Aurigma combination a try.

The photo I resized using Picasa to 1024x768 with preserved quality was uploaded using Aurigma with the following result (hold your breath):

Before: Size 262,005 bytes, WxH 1024x768 px, Res 96x96 dots per inch, Quality 94(411)

After: Size 218,258 bytes, WxH 1024x768 px, Res. 96x96 dots per inch, Quality 90(411)

Note the quality only drops slightly from 94 to 90 and their is NO DPI change! I'm guessing that the uploader doesn't need to RESIZE, which is possibly where the quality loss and/or the DPI change is made. Is that a winner or what?

So here's the best pic receipe.

1. Use Picasa to resize the photo to 1024x768 (or max dimension as applies). The resize is done using the "Export" feature of Picasa, which places the photo in an Export Folder (nice). It looks like this can even be done with batches!!!

2. Upload using the Aurigma uploader.

I hope someone can duplicate my results with problem photos. Please!

Danno

And, here's the stored winner.

Hi Danno-it's good to know that it wasn't just me that saw a difference in uploaded images!
I don't have the same issues with the rare photos that I upload.
Most of my images are scanned in 100 year old postcards which by their nature rarely if ever have the quality that we now get from digital photos-this means that I am already using poorish quality images [especialy the printed postcards which vary in print quality by an enormous degree!] even before I scan and resize and then have the result compressed when uploaded!
I usually scan my postcards at 200dpi into 'Irfanview' [free but excellent editor etc. well worth a try!]
I then resize to 1024 x 768 [or as close as possible] save them at 80% quality at 72dpi before uploading to Auctiva.
I'll save the next ones that I do at 100% and 96 dpi and 1024 and see if there is an improvment.
The next question that occurs to me is should the resizing of the image be done before saving at 96 dpi or after?

Hi,

My only thought on the DPI issue may be any dithering of the image due to "creation" of more image dots in combination with applied compression. Frankly, I saw only very very small differences in the uploaded photos I tested. Only inspection with Picasa and Paint told me there were real differences in the stored quality. A quality of 90 or higher should have very little artifacts, so sticking with the Aurigma uploader is a better choice for very high quaility images.

Bypassing the resize done by the Aurigma uploader might avoid any artifacts induced with that odd DPI change observed. My suggestion is to use Picasa (or other product with similar capability) to do the resize and preserve the quality factor at the level of the original image, then use the Auctiva/Aurigma Uploader (not the HTML).

I've been checking on that 411 in the quality number, e.g. 90(411). I believe that's the jpeg format number. I tried a test with my camera set to "Fine" and got a quality index of 95(422) from the raw 6.1MP image (and a larger file size). 422 is also a jpeg format number. I tried uploading the image without resize and the stored image was converted by Aurigma to 90(411). No artifacts or problems, but it does give me one more possible avenue to explore for the reported distortion problems. I'm aware there is a jpeg 111 format number, maybe more.

Danno

Edit: The jpeg number is apparently the sampling decimation factor, e.g. 411 is YUV 4:1:1.

quote:

I'm so confused.

Join the crowd.

Under normal circumstances, I'd agree that DPI "shouldn't" influence the quality. However, I'm curious about that factor in combination with the other changes I've noted:

1. change in compression

2. change in decimation ratio

3. resize sampling interpolation

All of those have the potential to induce artifacts. So far, I'm unable to find a formula that replicates reports with normal digital cam photos.

The subject of the thread was "optimum uploading size". My tests suggests you get the best quality by using your own resizing tool, (like Picasa) to size to the 1024W x 768H px limits and then use the Auctiva/Aurigma uploader. If that formula doens't work for those having artifact problems, I'd need to have a step-by-step on what they are doing with quality data to help replicate the problem.

As time permits, I'm going to try testing scanned images with my 1200 DPI scanner.

Danno