Here's a good lesson on source pict size versus excessive compression methods. Suppose that you have a fixed file size in mind for a finished panorama, say under 200k. What size images should you begin with? And which compression method should you use?

     For a web based panorama displayed in a standard 320 x 240 window, using the 768 x 512 image size should be the highest resolution necessary. Remember that these images are captured in portrait orientation, so the vertical size is 768 pixels, over three times the vertical size of the window (240 pixels), allowing for plenty of zooming data.

     The panorama shot in this example is an outdoor scene, capturing an autumn day in Chicago. I used Kodak Royal Gold negative film, rated ISO 100, and shot the images at f/16 @ 1/100 of a second. Using a 24mm Nikkor Lens, I captured 12 images, at 30° apart, allowing for enough overlap during stitching.

     Generally, I use QTVRAS as a two step process: first to just stitch the source pict, and second to build the QTVR file, after some post production in Photoshop. For the stitch, I turn off the sharpen function in QTVRAS, as I feel that performing that interactively in Photoshop yields better results. After QTVRAS finishes stitching the image, I have a file size of 740 x 3552, currently 7,885,440 bytes uncompressed.

     My post production on this image consists of creating a sharpened layer, a blurred layer, masked to reveal only the sky, and a levels adjustment layer. I aggressively sharpened this image to show the detail in the buildings. I generally mask & blur the sky on all outdoor shots, finding that this step usually creates smaller panos, but can producing gradient banding noticeable with Photo JPEG and certain other compressors.

     Back into QTVRAS, this time performing a "Make Panorama" with the modified Photoshop file (the ability of QTVRAS to use layered Photoshop files is one of it's greatest features). I turn off automatic tiling, manually setting the tiles to 1 by 12, because using less tiles should produce a more compressed file. My preferred compressor is Sorenson, though I expect it might not work well with this shot. Sorenson tends to shift green colors, and is not always visually appealing on daytime outdoor panos.

NOTE: To view these files you must have a JavaScript enabled browser. This will allow you to have multiple windows open to compare the differences in quality. When comparing files, zoom in all the way, and examine the man standing by the bench, or the van parked on the street. Notice the difference between compressors.

My first attempt with Sorenson set at 7% quality yields a file 261,773 bytes (30.12:1 compression ratio), slightly larger than I want. View file.

Lowering the Sorenson quality slider down to 3% provides me with a file 218,257 bytes (36.13:1 compression ratio), which is an acceptable size, however the overall image quality leaves much to be desired. View file.

I move on to Photo JPEG, generally the next best compressor for QTVR panoramas. With a quality setting of 5%, I get a file 263,533 bytes (29.92:1 compression ratio), which is still too large of a download. Notice that the detail is much higher than the Sorenson compressed files, however Photo JPEG compressed panos tend to play back with a noticable ripple effect on slower computers. View file.

Now for Cinepak, the original favorite for QTVR panoramas, before QuickTime 3 was released. At 10%, I'm able to achieve a file size of 223,845 bytes (35.23:1 compression ratio), which is an acceptable size. Unfortunately, I feel that too much detail has been lost, because compressing so heavily has compromised the image. The file takes on a painterly look, and is more pixelated. View file.

A rarely used codec is Intel's Indeo 5, which at 5% quality produces a file 273,789 bytes (28.80:1 compression ratio). The image quality is similar to Cinepak, except it will not compress as much. This is still larger than I wanted, so I conclude that the original source pict will have to be resized to a smaller image. View file. NOTE: To view this version, you must have installed the Intel Indeo 5 codec for QuickTime 3 or later. If you need it, get the free download from Apple at this page.

Back into Photoshop, I resize the image down to 480 x 2304, which should still provide enough zoom detail, yet at 3,317,760 bytes is about 42% the size of the original image. The advantage here is that Photoshop will preserve as much detail as possible, so when I use this image I will not need to compress as heavily. This gives me more of an accurate idea before hand as to how the final movie will appear, instead of any unexpected smearing of colors due to excessive compression.

Back to Sorenson, I set the quality to 25% and end up with a file 208,945 bytes (15.88:1 compression ratio). This is close enough to the 200k target size, but I still want to see if Photo JPEG will retain the colors more accurately. View file.

With the quality of Photo JPEG set at 34%, I get a file 205,950 bytes (16.11:1 compression ratio). This file is just slightly smaller than the Sorenson version, but the colors are better, so this looks like it's the one I'll use. It's also worth noting that the overall playback of the movie using the smaller source image will be more responsive on older machines, and memory management will be better. The disadvantage to using Photo JPEG is that the sky will appear as a gradient banded area, whereas using Sorenson, with it's built-in bicubic interpolation, smoothes out these continuous tone areas. View file.

     Hopefully this has given you some better ideas on compressing panoramas. Unfortunately, every image is different, and requires different methods. It's important to have a planned method of attack, determine what is most crucial, and work to reach that goal.