Adaptive Color Lookup Table for GMM Video Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Manual segmentation of video images into foreground and background is time-consuming, tedious, and prone to errors, making it inefficient for digital video applications.

Innovation Solution

The application of Gaussian Mixture Models (GMMs) to local image patches using an adaptive color lookup table, which performs per-channel color quantization to find representative colors and computes probabilities for each pixel, allowing for automated segmentation by propagating initial segmentation information across frames.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual segmentation is used to separate foreground and background in video images, then segmentation accuracy can be maintained through human judgment, but the process becomes time-consuming and tedious

Engineering Contradiction:

Improvesegmentation accuracyVSAvoidsegmentation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system enables automatic segmentation by having the computer system perform the segmentation task itself using GMM and lookup tables, eliminating the need for manual human intervention while maintaining accuracy through algorithmic color analysis and probability computation

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical manual process of visual inspection and manual segmentation with an automated computational system using Gaussian Mixture Models and color lookup tables to perform segmentation automatically

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If automated segmentation methods are used to reduce time consumption, then processing speed improves, but accuracy may deteriorate compared to manual segmentation

Engineering Contradiction:

Improvesegmentation speedVSAvoidsegmentation accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent uses per-channel color quantization to transform continuous color values into discrete representative colors, and adjusts probability thresholds to optimize the balance between processing speed and segmentation accuracy for automated operation

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces an adaptive color lookup table as an intermediary structure that pre-computes and stores probability values for representative colors, enabling fast automated segmentation while maintaining accuracy by referencing pre-analyzed color probabilities

Inventive Principle:

Principle #24Intermediary (Mediator)

3Measurement precision

If per-pixel GMM computation is performed for all pixels in image patches, then accurate probability computation is achieved, but computational time increases significantly

Engineering Contradiction:

Improveprobability computation accuracyVSAvoidcomputational time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent divides the image into local patches and applies per-channel color quantization to reduce the number of unique colors within each patch, thereby reducing the number of GMM computations needed while maintaining local accuracy

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent pre-computes and stores probability values in an adaptive color lookup table for representative colors before processing actual image pixels, eliminating redundant GMM computations and significantly reducing processing time

Inventive Principle:

Principle #10Preliminary action

4Speed

If a large color lookup table is used to store probabilities for all possible colors, then color probability lookup is fast, but memory requirements increase

Engineering Contradiction:

Improveprobability lookup speedVSAvoidmemory usage

Core Design Contradiction:

SpeedVSQuantity of substance

Solution Approach 1:

The patent applies per-channel color quantization to reduce the number of color levels from 256 per channel to a smaller number of representative colors (e.g., 16-32), dramatically reducing lookup table size from 256³ to m³ entries while maintaining color representation quality

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent creates adaptive lookup tables for each local image patch based on the specific color distribution in that region, storing only the probabilities for representative colors actually present in each patch rather than all possible colors

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS8290253B1Method and apparatus for applying Gaussian Mixture Models to local image patches using an adaptive color lookup table

Publication Date: 2012.10.16 ADOBE INC

AI summary

A method, system, and computer-readable storage medium for applying Gaussian Mixture Models (GMMs) to local image patches using an adaptive color lookup table. Per-channel color quantization may be performed to find representative colors for a local image patch. Each combination of the representative values corresponds to a representative color. The probabilities of the representative colors may be computed using a local GMM color model and stored to corresponding entries in an adaptive color lookup table. For every pixel in an image patch, the closest representative color may be found, and the corresponding probability may be retrieved from the lookup table and used for the pixel. The method may, for example, be applied to each local window in a method for automatically determining segmentation in a digital video image sequence to calculate the foreground probabilities for the pixels in a propagated classifier via a GMM.