ImageVice™ is a new, proprietary color reduction technology, which makes possible greatly improved compression ratios for all indexed file formats employing lossless compression.
GIF, PNG, PICT and BMP are the most common graphics formats that can benefit from ImageVice™ color reduction technology.
Applications for Web and multimedia graphics are tremendous. With ImageVice™ color reduction smaller graphics files can be produced and smaller files translate directly to faster loading web pages or more content possible for Web and multimedia within the available storage constraints.
Lossless compression techniques such as Lempel-Ziv Welch "LZW", which the GIF format uses, and Run-length Encoding "RLE", which is used by both the PICT and BMP formats, is not variable and can not be made significantly more efficient.
Given the same data to compress, two LZW compressors will make the exact same size compressed file.
The only variable in the final file size for indexed color graphics in GIF, PNG, PICT or BMP formats is the initial color reduction step. It is that step that determines the actual image data to compress.
The normal approach taken by color reduction algorithms is to create the most mathematically accurate data to represent an image with a limited palette of colors. This approach is perfectly fine when there are no considerations needed for final file sizes, but in applications such as Web and multimedia graphics file size is critical. It effects both speed and content capacity. The most mathematically accurate data is the least compressible data in the case of color reduction.
Human visual perception is not a mathematical process, though. The perceptual criteria of the human visual system is very different to the algorithmic processes of normal color reduction techniques.
ImageVice™ technology takes advantage of the perceptual processes of human vision to model image data in such a way that it is not only visually accurate but also much more compressible than otherwise.
The processes used in ImageVice™ technology can be controlled to a great extent providing broad capability to produce images which are more visually accurate or more compressible as needed, another advantage of ImageVice™ technology not provided by other methods.
At the high end, image quality is much better than all other known color reduction methods. Under normal usage settings ImageVice™ technology will exceed other color reduction methods in both quality and compressibility of image data produced.
Using extremely aggressive settings for compression will result in the smallest possible files, but will begin to degrade image quality in noticeable but not usually objectionable ways. Images produced with aggressive settings for compressibility will have a painterly look to them.
There are indeed many comparisons between the process employed by ImageVice™ and the process a painter would use to represent a scene as a realistically recognizable image.
Installing the ImageVice™ plug-in:
ImageVice™ is a Photoshop filter plug-in that is compatible with Photoshop versions 3.0 and greater.
To install the ImageVice™ plug-in, place the plug-in, the file named "ImageVice™ 1.0", in your Photoshop "Plug-Ins" folder located within Photoshop's application folder. You may also place the "ImageVice™ 1.0" plug-in inside the "Filters" folder within the "Plug-ins" folder, but this is not required.
The next time you restart Photoshop, the ImageVice™ plug-in will be available for you to use. It will appear under the "BoxTop Web Tools" sub-menu, which will be added to Photoshop's "Filters" menu after the ImageVice™ plug-in has been installed.
Basic usage of the ImageVice™ plug-in:
The ImageVice™ plug-in is a filter plug-in that acts on RGB mode images, and is not tied to any particular graphics file format. This provides greater flexibility for use.
To use the ImageVice™ plug-in, you first run the filter on your RGB image. This changes the image to contain 256 or fewer colors. Then you change the document mode in Photoshop to "indexed" color with the defaults that will appear in the mode change dialog.
After you have changed to indexed mode, the image can be saved in the normal manner that you would save GIF, PNG, PICT or BMP files.
For GIF graphics, you should use ImageVice™ in combination with a specialized production tool like PhotoGIF™, which provides the best format specific capability for making GIF files including more color reduction flexibility when you need to use an existing palette such as the Web safe palette for a particular image.
ImageVice™ can only create adaptive palettes based on the original colors in the image.
Understanding the ImageVice™ controls:
The ImageVice™ plug-in has a single, relatively simple, dialog for specifying color reduction parameters. Though, there are only six individual parameters, which can be specified for color reduction, not all of these are immediately obvious in their function.
The only controls in ImageVice™ that are comparable to controls found in other color reduction tools are the depth menu and colors text box. Both of these provide a means to set the number of colors to be used in the final image.
All other controls are unique to the ImageVice™ plug-in, and require a level of understanding to get the maximum benefits possible from the ImageVice™ plug-in.
Balloon help is available in the ImageVice™ dialog to help understand the function of the various controls and overcome the small learning curve.
More detailed explanation for controls are provided below.
1: Depth menu
The depth menu shows the lowest possible bit depth for the image given the current number of colors specified, and can also be used to set the number of colors specified to the maximum number of colors possible for a given bit depth.
Lowering the bit depth will improve an image's compressibility and reduce the final file size. Very few images need to have more than 128 colors, which is the maximum for a 7 bit image, to adequately represent them and many image require even fewer colors.
2: Colors text box
The colors text box provides another means to set the number of colors to be used. Odd numbers of colors can be specified using the colors text box, whereas only even bit depths can be specified using the depth menu.
Even if lowering the set number of colors will not allow a lower bit depth, it will still improve compressibility. An image using 120 colors will normally be smaller than an image using 128 colors, though both images are 7 bit.
You should set the colors to the minimum number that will produce satisfactory visual results for the individual image. If you are unsure of what this number is, starting at 128 colors or 7 bit, is a good idea. Then you can undo the filter and try a higher or lower number if it looks like fewer colors will do or more colors are needed.
3: Clipping sliders
The clipping sliders are a set of two slider controls that are linked together in their function.
The first slider is lower or dark color clipping. The second slider is upper or light color clipping.
The lower clipping slider controls the threshold level for sampling dark colors from the image. The range of the lower clipping slider corresponds directly to gray component values 0 - 50. If the lower clipping slider was set to 50, no colors would be sampled from the image that had a gray value lower or darker than 50 during color reduction.
The upper clipping slider controls the threshold level for sampling light colors from the image. The range of the upper clipping slider is scaled to correspond to gray component values 236 - 256 inversely, which is best thought of as the number of high or light gray levels not to sample. Setting the upper clipping slider to 20, has the effect of no light colors with a gray values of 236 or above being sampled during color reduction.
Mid range values for clipping will improve image compressibility by mapping all colors that fall below the thresholding values to a single color in the ranges subtleties in color are most difficult to distinguish and reducing entropy. Image quality will also be improved with mid range clipping settings because more palette entries will be available for mid range colors that are more easily distinguishable.
You should reduce the lower or dark clipping setting for images that are predominantly very dark colors to prevent harsh borders where thresholding occurs and leave more palette entries available when they are actually needed in the image.
Correspondingly, you should also reduce the upper clipping setting in images that are predominantly very light colors for the same factors.
In images that have few dark or light areas, you can safely increase the settings for clipping to very high values to improve compression.
4: Smoothing slider
The smoothing slider controls the tolerance for when very similar adjacent pixel color values are merged.
This is a complex space variant process with only one controllable parameter, so the action taken will vary significantly depending on the actual image, unlike the upper and lower clipping sliders which have predictable results on all images.
Increasing the value for smoothing will increase the compressibility of the image and normally has little apparent visual impact on the final image. However, if the original image has predominant areas that are very smooth linear color blends, high smoothing settings will become more obvious and possibly undesirable.
5: Convergence sliders
The two sliders for convergence work together to specify a ratio for an internal mechanism that there is no sane way to explain.
The first slider is an overall tolerance value for this mechanism and the second slider provides a measure of relational amplitude.
Though, the process of convergence is not easily explained in words, the visual effects are apparent and their nature is easily understood from visual example.
To better understand the visual effects of convergence settings, you should compare images filtered with both convergence sliders set to zero to an image filtered with default convergence values and then one filtered with both convergence sliders at maximum.
Convergence settings are primarily responsible for both incredible compression levels that can be achieved with the ImageVice™ plug-in and at the highest settings the aforementioned "painterly" look to images.
The default settings for convergence work very well with almost all images, and produce very good quality to compressibility ratios.
Increasing the first convergence slider value will have less visually apparent effect than the second, which should rarely be used at its maximum setting. However, when maximum compressibility is the priority, convergence settings should be increased first as they have the most impact on compressibility.
When maximum quality is desired the convergence sliders can be set to values lower than the defaults, but it is not recommended for compressibility's sake to turn convergence off completely. Very low settings will still significantly improve compressibility without any noticeable visual effect on the image.
6: Image preview
The image preview is only thumbnail size, which is less than perfect for the purpose of color reduction, but the most that can practically be provided given the computational complexity involved. Of all the great things that can be said about the ImageVice™ plug-in, "fast" is not one of them.
The image preview does provide some amount of quick feedback on chosen settings, but it should be realized that the thumbnail isn't as high in quality as the final image will be because of the scale factor.
7: OK and Cancel
Both do the usual thing.
Trying out the ImageVice™ plug-in:
The ImageVice™ plug-in demo is fully functional, but will only work with the six provided test images, which it calculates a checksum to identify. These six images were chosen very carefully to be the most representative test suite for all types of images from photographs to rendered graphics. They are "combination" images, artificially made with elements from all types of images.
You will be able to experiment with the ImageVice™ plug-in on the images provided, and compare the results to other tools. Take care not to save over the original test images. Any modification to them and the ImageVice™ plug-in will no longer work on them.
When comparing file sizes you should use a utility such as ResEdit, and its "Get File/Folder Info..." command that shows the size of the file's data fork separately from the size of the resource fork.
The data fork contains the actual graphics file. Any information in the resource for is platform and host specific esoterica, such as the last window position of your working document, custom icons and previews.
The Macintosh Finder's Get Info dialog shows "Size: 182K on disk (173,412 bytes)" for example. The first number in whole K is the allocation size for the file, or the amount of hard disk space that is assigned to hold the file. The second number is more accurate than the first. It is the bytes used, or actual combined size of the data and resource forks.
Depending on what is actually in the resource fork, the bytes used number may or may not be representative of the actual data size.
The only way to know for sure is to use a utility such as ResEdit, which shows the size of the data and resource forks as separate numbers. ResEdit is available as freeware from Apple Computer's web site at http://www.apple.com.
If you have any questions while evaluating the ImageVice™ plug-in demo that this documentation does not answer, please let us know.
Contact information is below.
BoxTop Software, Inc.
101 N. Lafeyette Street
Starkville, MS 39759
(v) 1-601-324-1800
(f) 1-601-324-7352
http://www.boxtopsoft.com
ftp://ftp.boxtopsoft.com/pub
info@boxtopsoft.com
Sales contact:
Cindy Johnson - cindy@boxtopsoft.com
Support Contact:
Susan Brewer - susan@boxtopsoft.com
Company Contact:
Travis Anton - travis@boxtopsoft.com
Development Contact:
Richard Anton - ranton@boxtopsoft.com
ImageVice™ 1.0 is Copyright 1997 BoxTop Software, Inc. All Rights Reserved. ImageVice™ and PhotoGIF™ are Trademarks of BoxTop Software, Inc. All other trademarks are recognized as the properties of their respective holders.
Document published: 11/10/97
Copyright 1997 BoxTop Software, Inc. All Rights Reserved.
