Anisotropic Filter Array for Scale Invariant Video Fingerprinting

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video processing technologies face challenges in accurately identifying and indexing large video databases due to issues with image distortion, geometric and optical transformations, and the need for scalable solutions that can handle high-definition video formats efficiently.

Innovation Solution

A method involving a two-pass analysis using anisotropic filters for detecting interest points and generating multi-dimensional signatures, which are resistant to image distortion, by applying elliptic and rectangular-shaped sampled Gaussian filters for scale-space analysis and feature extraction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If complex filters are applied to all video frames for accurate region characterization, then measurement precision is improved, but device complexity and processing time increase

Engineering Contradiction:
Improveregion characterization accuracyVSAvoidfilter processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the video processing into two distinct passes: a first pass using simple filters to identify candidate interest regions, and a second pass using complex anisotropic filters only on those identified regions. This segmentation allows complex processing to be applied selectively rather than universally, reducing overall computational complexity while maintaining measurement precision for regions that require it.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different filtering strategies to different regions of the video frames based on their characteristics. Simple filters are used for initial screening of all frames, while complex anisotropic filters are applied only to identified interest regions where high precision is needed. This local differentiation of processing quality optimizes the balance between accuracy and computational cost.

Inventive Principle:
Principle #3Local quality

2Manufacturing precision

If complex filters are applied to all video frames, then manufacturing precision of feature detection is improved, but productivity decreases

Engineering Contradiction:
Improvefeature detection accuracyVSAvoidvideo frame processing throughput
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The processing workflow is segmented into two stages: rapid initial detection using simple filters across all frames, followed by detailed analysis using complex filters only on candidate regions. This segmentation enables high throughput in the first pass while ensuring high precision in the second pass for critical regions, thereby maintaining both productivity and feature detection accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Instead of applying complex filters to all frames (excessive action), the patent applies them only to a subset of identified interest regions (partial action). This partial application of complex processing is sufficient to achieve the required feature detection precision while avoiding the excessive computational burden that would reduce productivity.

Inventive Principle:
Principle #16Partial or excessive action

3Adaptability or versatility

If scale invariant interest region detection is implemented, then adaptability to image distortion is improved, but device complexity increases

Engineering Contradiction:
Improveresistance to image distortionVSAvoidfilter array complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements scale invariant detection through a two-pass approach where the first pass identifies candidate regions using simple scale-space analysis, and the second pass refines these regions using an array of anisotropic filters at multiple scales. This segmentation of scale analysis across passes reduces the complexity burden of full scale invariance while maintaining adaptability to distortion in the final result.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extends the filter analysis from a single scale to multiple scales by implementing an array of anisotropic filters with varying scale parameters. This dimensional extension into the scale domain enables the system to detect interest regions that are invariant to scaling and distortion, achieving high adaptability while the multi-scale approach is applied selectively to identified regions rather than uniformly.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Data Source

PatentUS9652672B2Digital video content fingerprinting based on scale invariant interest region detection with an array of anisotropic filters
Publication Date: 2017.05.16 ROKU INC
  • US9652672B2 patent drawing
  • US9652672B2 patent drawing
  • US9652672B2 patent drawing

AI summary

Video sequence processing is described with various filtering rules applied to extract dominant features for content based video sequence identification. Active regions are determined in video frames of a video sequence. Video frames are selected in response to temporal statistical characteristics of the determined active regions. A two pass analysis is used to detect a set of initial interest points and interest regions in the selected video frames to reduce the effective area of images that are refined by complex filters that provide accurate region characterizations resistant to image distortion for identification of the video frames in the video sequence. Extracted features and descriptors are robust with respect to image scaling, aspect ratio change, rotation, camera viewpoint change, illumination and contrast change, video compression/decompression artifacts and noise. Compact, representative signatures are generated for video sequences to provide effective query video matching and retrieval in a large video database.