Developmental History and Projections

Data Tagger: 2/86 - 8/99    The data tagger transforms the disconnected dots into primitive mathematical tags which allows for better electronic designs. It was started in 2/86 and finished in 8/99. The data tagger needs more testing as the glider is completed. Glider Convolver: 2/86 - 4/02    The glider convolver transforms the primitive mathematical tags into the desired program output. It is designed to be a toolbox that allows other developers to transform our mathematics into their desired output. It was started in 2/86, and it still needs debug. We intend to finish the glider convolver after the blob compressor and color segmenter are complete. Color Segmenter: 8/99 - 6/01    The color segmenter groups colors together at photographic quality which is used as a noise reduction technique. At the same, it allows for better compression. The segmenter was started in 8/99 and completed in 6/01. The settings of the color segmenter need to be manually set, and we are currently building an auto-set feature that will remove this requirement. Blob Compressor: 6/01 - 8/10 The blob compressor groups shapes together which is used as a noise reduction technique. At the same, it allows for better compression. The blob compressor was started in 6/01 and passed all engineering tests in 10/06. As a large piece of code that had no precedent, some effort was required to reach an industrial version. The Pac-n-Zoom Pre.Vu version of the blob compressor was placed on the Internet for download in 3/09, but it is still somewhat instable. We dropped off the blob compressor in 8/09 to concentrate on the auto-setting segmenter, but we will return to the blob compressor to make it stable after the auto-setting segmenter is released. 3/09: Pac-n-Zoom® Pre.Vu    The initial release allows interested parties to try out the software on black and white files (no colors or grays). It will cleanup FAXes, recognized text, collate documents, and a few other convenient things. The initial release will have the following structure.

Auto-Set Color Segmenter: 5/09 - 6/10    The color segmenter groups colors together at photographic quality which is used as a noise reduction technique. At the same, it allows for better compression. The segmenter that was started in 8/99 and completed in 6/01 required manual settings which limited commercial viability. This version of the color segmenter has an auto-set feature that removes the manual requirement and allows for full commercial utility. 6/10: Pac-n-Zoom Color    This product is Pac-n-Zoom Pre.Vu with the automatically set color segmenter which removes much of the noise from an image. This process allows PNG to compress about as much JPG. We are testing whether it will allow a wavelet video compressor (e.g., Dirac) to offer superior compression over a DCT video compressor (e.g., like that usually associated with MPEG).

8/10: Pac-n-Zoom Raster Video    This release of Pac-n-Zoom will have a streaming (video over the Internet) video compression mode. We think the compression will be about 100 times better than what the current standards do, and the quality of the video should be better than the original video. Video over the Internet, however, has many variables, and without further testing with more developed code, we are not sure of its commercial viability. For example, we are not sure that the computer can decode the Pac-n-Zoom file fast enough to support video speeds. On the other hand, the initial calculations indicate it will work fine. The video block diagram is a little more complicated than the color release. 11/10: Pac-n-Zoom ASP    The ASP release is the same as the video streaming release but with a twist. The ASP is used to lower the cost of computing while providing many other computing advantages (see ASP White Paper). Since computer video doesn't usually move very fast in comparison to TV video, the ASP video compression can be enormous, and even a dial-up might work in some cases. Pac-n-Zoom provides the video compression needed to make ASP a viable technology. The following diagram shows how Pac-n-Zoom fits into an ASP design.

Click on a specific area for more information

User: In the ASP model, the user can be far away from the computer that is running the application. In fact, the user may be operating applications in 6 different continents at the same time without knowing it. Keyboard: This is the user's keyboard. Mouse: This is the user's mouse. Monitor: This is the user's monitor. Windows Messaging: This is a software module that resides inside the operating system of the user's computer. Among other types of information, the messaging system picks up the mouse and keyboard data. Windows Bitmap: This is a software module that resides inside the operating system of the user's computer. The Window's operating system displays the image inside the Windows bitmap memory. Data Packer: This software module resides on the user's machine, and we are using it to make the mouse and keyboard data compatible with Internet. Pac-n-Zoom Decoder: The Pac-n-Zoom viewer resides on the user's machine. Data Unpacker: This software module resides on the computer that serves the application. It makes the data transmitted over Internet look like it comes from the keyboard and mouse. The software would be provided by a third party. Pac-n-Zoom Encoder: The Pac-n-Zoom encoder resides on the application server and provides impressive video compression that makes the ASP model viable. Windows Messaging: This is a software module that resides inside the operating system of the application server. The user's keyboard and mouse data are inserted into the operating system. Windows Bitmap: This is a software module that resides inside the operating system of the application server. In a conventional system, the Window's operating system displays the image inside the Windows bitmap memory, but in this case the image is sent over the Internet before it is displayed. Application: This is the software that the user wants to run. Nearly any application that can be run on a personal computer can be run with this model or one that is very similar. For example, a driving program might have a steering wheel in addition to the keyboard and mouse. In this case, the steering wheel is installed into the user's computer in the same way it would be installed in the conventional computer model. 3/11: Pac-n-Zoom Raster Audio    This release of Pac-n-Zoom will have a streaming (audio over the Internet) audio compression mode. We think the compression will be about 10 times better than what the current standards do, and the quality of the audio should be better than the original audio. 4/11: Pac-n-Zoom Math    With the mathematical release, we will begin to introduce our core technology. The glider resolver tool set will not be either highly developed or well understood, but we will be able to do resolution enhancement which is a needed solution in graphic arts, motion pictures, and other areas. We will open source the glider resolver, GUI wrapper, and process controller. As the block diagram shows, this release coincides with demo 8 which is probably our most significant milestone.

Match Extremities: When the tags reach the glider they are formated as significant points on a blob surface. Since a blob requires two or more surfaces, the surfaces need to be matched. The surface extremities are used to match the surface. Assign Quadrants: After all the tags of a specific blob are known, each tag is assigned a quadrant with respect to the centroid of the blob's area. Assign Slopes: The assignment of the slopes ensures a continuous mathematical function between equations. Since there are so many possible equations, there is much more to this step than what it might seem. The tags are encoded in three different ways.

1. Relative Positions: These positions include the adjacent tags and their adjacent tags. 2. Sequential Orders: These include quadrants, positions, and tag types. 3. Tag Types: The tag types of up to 5 sequential tags are used to define the slopes. There 4 basic tag types.

After the slopes are assigned, the each tag contains the infomation to either build a library tag or be reduced to raster for display. Insert Turns: The remaining steps involve the reduction of tags to raster. In other words, if the insert turns step is taken, raster is the only possible output. This step places other tags into the sequence as required by raster. Build Steps: The steps are a specific number of points between two tags. Match Grid: This probably is the part of the glider that needs debug. The grid needs to be matched in build steps. Build Bitmap: The Windows bitmap is written, and the operating system takes it from here. Only the pixels that change from either the last line or frame, depending on the application, are updated. 7/11: Pac-n-Zoom OCR    With the OCR release, we want to show solution providers how to implement the tool box. The convolver develops the specified solution with input from the golden library which is an outside database. Each cycle of the convolver adds another level of hierarchical structure to the solution. In the OCR example, the following would match on each iteration.

1. Single Blobs: In the initial iteration, the individual blobs are matched to the golden library on the first iteration (e.g., the inside island and perimeter of the letter 'e'). 2. Single Characters: In the second iteration, characters, such as alphanumeric characters are matched (e.g., the letter 'e'). 3. Two Characters: During the third iteration, two character strings are matched (e.g., the character sequence "he"). 4. Three Characters: Three character strings are matched during the fourth iteration, (e.g., the character sequence "The"). 5. Four Characters: Four character strings are matched (e.g., the character sequence "They") in the fifth iteration. 6. Single Words: The interations continue until the longest words are matched. Then word sequences are matched. (e.g., the sequence "They will"). 7. Single Paragraphs: The interations continue until all the word sequences are matched. For example a five word sequence would require 5 iterations. Then paragraphs are matched. 8. Single Pages: The interations continue until all the paragraphs are matched. Then the pages are matched.

While this example was for text, the convolver will work with any set of shapes. For example, the lines of a form could be matched just as easily as the text. The following is a rather involved diagram of how the convolver is implemented into the glider.

Select Library: Since the convolver will be used in different applications, the application specific library needs to be selected. In this text recognition case, for example, we would choose the text library, but we could choose other libraries as well. All selected libraries will run simultaneously. Correlate Library: The correlator matches any golden tags and places the tags in the cyclic convolving data stream to build hierarchical data tags. The convolving cycle continues until the convolving produces no further hierarchy and no golden tag is correlated within an entire cycle as in the example given above. Library Tags: The most useful set of library tags is an exhaustive set of tags at a single level of hierarchy. For example, every character of every font in an OCR application. Multiple levels of hierarchy can be used for compression but not for correlation. Golden Library: The golden library is an external database that contains the library tags for specific applications. For example, voice recognition would use a different set of tags than OCR. These tags are loaded into glider before the application begins. Select Output: Pac-n-Zoom should eventually become compatible with many other applications. This selection allows the user to specify the output(s). Build File(s): This code, which is written by a third party, creates a file that is compatible with another application. Application File(s): These files are compatible with third party applications such as drawing, simulation, CAD, or animation programs. 9/11: Pac-n-Zoom Vector Audio    Before this release, Pac-n-Zoom only turned two dimensional input data into mathematics, although some output applications could be three dimensional. In this release, Pac-n-Zoom will compress single dimensional data which is the kind of data that most sensors and applications require. Since audio is the predominant single dimensional data application, the release is called audio, but other single dimensional data applications (e.g., ultrasound, microwaves, stress, etc. or even database streams such as day to day prices) are also supported. TBD: Pac-n-Zoom Vector Video    With this release, Pac-n-Zoom will be extended to three dimensional input data which will be horizontal, vertical, and time. The primary application is video, but there are other applications such as remote sensing radar where time is related to the distance of the artifacts.