Terrain Map Summary Elements for Hierarchical Image Abstraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image understanding systems lack the ability to efficiently create higher-level abstractions from pixel data for machine vision, limiting their capability to analyze and understand complex image content effectively.
Innovation Solution
The development of a hierarchical tree of symbolic data structures, known as Terrain Map Summary Elements, which summarize lower-level data to provide richer, higher-level descriptions of image content, enabling more efficient machine vision analysis and image retrieval from large databases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If hierarchical Terrain Map Summary Elements are created to provide richer image content descriptions, then image understanding capability is improved, but data processing complexity increases
Solution Approach 1:
The patent segments image data processing into multiple hierarchical levels: raw pixel data is divided into Terrain Map elements (4 pixels per element), which are then grouped into Terrain Map Summary Elements (16 lower-level elements per summary element). This segmentation allows complex image understanding tasks to be broken down into manageable processing stages, improving capability while controlling complexity through structured organization.
Solution Approach 2:
The patent introduces a hierarchical dimension to traditional image processing by creating multiple levels of abstraction from pixels to Terrain Map elements to Summary Elements. This dimensional transformation enables the system to handle complex image understanding requirements by operating at different levels of the hierarchy, selecting the appropriate level based on task requirements to balance capability and complexity.
2Adaptability or versatility
If multiple levels of abstraction are created from pixel data, then higher-level image understanding is achieved, but data bandwidth requirements increase
Solution Approach 1:
The patent merges multiple lower-level Terrain Map elements into higher-level Summary Elements, where each Summary Element aggregates information from 16 lower-level elements. This merging reduces the total number of data structures that need to be processed and transmitted, as the hierarchical structure allows operations to be performed at the summary level rather than requiring all detailed pixel-level data to be handled simultaneously, thus controlling bandwidth requirements.
3Measurement precision
If Terrain Map elements process 4 pixels per element with 8 primitives each, then detailed image analysis is enabled, but processing time increases
Solution Approach 1:
The patent performs preliminary processing by creating Terrain Map elements from raw pixels first, extracting 8 primitives (AverageAltitude, DegreeOfSlope, DirectionOfSlope, HorizontalSmoothness, VerticalSmoothness, Jaggyness, DegreeOfColor, DirectionOfColor) for each element. This preliminary action prepares data in advance at a manageable granularity level, so that subsequent higher-level analysis operations can work with pre-processed Summary Elements rather than raw pixels, reducing overall processing time while maintaining precision.
Data Source
AI summary
Method and system for image data analysis, by making use of a Terrain Map Summary Element (TMSE) data structure, in which a Terrain Map data structure provides processed primitive data about each 2×2 kernel of pixels from an original image. The new TMSE is a hierarchical multiple-level or tiered structure of such Terrain Map elements, in which each successive Terrain Map level summarizes, for example, 16 lower level elements or member. By providing such a hierarchical nature of the map elements, richness is enhanced by each higher element because every element of each successively higher level map summarizes the data from elements of the lower level map. For example, in a five-level Terrain Map Summary Element structure according to the disclosure, each element or member of the fifth level contains information summarizing data about 256K pixels.