Theory of Active Perception
Computer Vision
CV (for video)
Video indexing
Further development
Theory of Active Perception
All the existing methods and technical solutions for information processing deal with a narrow range of applications. Processing and decision-making normally take much longer than required. Segmentation and attributes generation stages have no theoretical basis yet. Without solving the problems of these stages, it is impossible to provide an optimal solution for information classification. TAPe copes with all those tasks. The Theory is based on a mathematical model that relies on group theory. TAPe uses a finite number of so-called elements-filters that, according to certain mathematical laws, are pooled into groups at three different levels. It is with their help that any information is recognized. Importantly, there is a minimally sufficient number of the first-level, second-level, and third-level elements, meaning they make up the exact amount—no more and no less—necessary to cope with any recognition task. And this helps tackle the issue of processing and decision-making speed, as well as deliver universal solutions.
But TAPe, as a new information processing method,
Сan be applied outside computer vision technology as well. New principles of building both neural network and computer processor architecture might be possible.
Computer Vision
Based on TAPe, we have developed a new information technology that can process any images by modeling natural mechanisms of human visual perception.

Thus, the human visual analyzer (the eye) has a number of objective properties ensured by innate mechanisms. Those mechanisms successfully tackle issues of recognizing any image in real time reliably and quickly even in the presence of any kind of interference, including when a bigger part of an object is obstructed. Thanks to them, people can determine what exactly they are looking at even if the object is only partly visible.

The core content of initial description elaboration stage for technology modeling (innate) visual perception mechanisms is processing any image right away as a whole. Thus, the main and only procedure does not involve convolution operation. As a result, there is zero correlation with noise (interference).

Integrity of perception
The process of analyzing an image is carried out as an integral (indivisible) systemic whole.
Swiftness of perception
The process of recognizing is carried out right away, in large blocks, and quickly, using a limited number of features.
Managing patterns
A pattern is a set of structurally related elements as compared to unrelated numbers that regular computers currently manage.
Due to TAPe, the system can recognize any image in any frequency range as fast as you can only imagine. Like the human brain, the system recognizes an image based on only a few key features.
Search by video
Video search and comparison technology using videos that is not conditional upon audio tracks, watermark usage, hash functions, text analysis etc.
The TAPe-based technology allows for both compact and distortion-resistant video indexes irrespectively of the original video stream format and type. It does not require advanced hardware and enables users to build, compare, and search by index in real time and with high accuracy using regular equipment. All of that provides effective solutions for a large task class involving media asset management, as well as asset protection and monitoring.
Requires a minimum of computing resources and traffic volume: 1 MB of traffic to send 1 hour of video over the Internet
Allows working with heavily noised images, different screen frames, side ratios, and frame rates
Search for 1 minute of video in 24 hours of a video recording happens in an instant—and that on a regular computer processor
Video indexing
Digital video indexing is a technology that allows identifying and coding content in the same manner as the human perception does. The system analyzes video content and determines the unique features that characterize it explicitly. Those unique features are digital indexes. They can be compared to the human DNA that also explicitly characterizes a particular individual.
Innovative video comparison and search algorithm using indexes
The resulting indexes are kept in the central database and can be further used to identify video data (files or streams) at its different lifecycle stages. To compare and search the requested video in the database, an innovative TAPe-based algorithm is used that determines the "visual similarity" between frames with the help of digital indexes. The algorithm is characterized with high speed and accuracy.
Helped by TAPe technology, the system builds digital indexes in a matter of milliseconds.
Index size
The list of attributes is compact and only takes up a few bytes per frame. Building an index for a sequence of frames allows for an even smaller size.
Using TAPe technology help ensure resistance to all kinds of interference and distortions that are inherent to both analog and digital signals—different noises, changes in brightness, color, contrast, macroblocking, scaling, shifting, etc.
Further development
Artificial intelligence and machine learning are not the limit. On the contrary, this technology may soon exhaust its possibilities. TAPe offers a new way to process information that models how human innate perception mechanisms work.

The theory is based on a mathematical model that replaces the binary number system. The new way to process information can be applied far beyond computer vision technology and can usher in new structural and architectural principles for neural systems and computer processors. For example, the new computer type is going to manage patterns rather than arrays of structurally unrelated numbers as it happens now.
A pattern is
a description using a subset of maximally informative related information elements. Patterns describing very different types of information (images, text, videos, sound) are identical, that is there is no differentiation according to information type. That is why for any task class, the amount of computing to be done will be reduced by an order of magnitude.
Made on