Looping Video Generation Using Edge-Based Transition Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional techniques for creating looping videos often result in unnatural transitions, such as teleportation of objects, and require high processing power due to the need to analyze all pixels in all frames, leading to inefficient and energy-intensive processes.

Innovation Solution

The method involves converting high-resolution video clips to lower-resolution versions, detecting edges, forming confusion matrices, filtering them, and identifying candidate transition points to create seamless looping videos by aligning start and end frames with similar edge features, thereby reducing global effects and prioritizing local alignment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional techniques analyze all pixels in all frames to create looping videos, then the accuracy of transition point identification is improved, but the processing power and energy consumption increase significantly

Engineering Contradiction:

Improvetransition point identification accuracyVSAvoidprocessing energy consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent extracts only the edge information from video frames rather than processing all pixels. By converting full frames to edgemaps that contain only edge pixel data, the system reduces the amount of data to be analyzed while preserving the critical features needed for transition point identification, thereby reducing processing energy consumption without sacrificing accuracy

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies local quality by focusing processing only on edge regions of frames rather than uniformly processing all pixels. The edgemap representation concentrates computational resources on boundary features where transition information is most relevant, eliminating wasted processing on non-edge areas while maintaining transition detection accuracy

Inventive Principle:

Principle #3Local quality

2Manufacturing precision

If conventional techniques process high-resolution video frames directly, then the precision of edge detection is improved, but the processing time and computational complexity increase

Engineering Contradiction:

Improveedge detection precisionVSAvoidprocessing time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent creates a simplified copy of the original high-resolution frames in the form of edgemaps. These edgemaps are lower-resolution representations that contain only edge information, serving as a proxy for the full frames. This copying approach allows rapid processing of transition points while preserving the essential edge features needed for accurate loop detection

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent segments the video processing task into two stages: first, convert frames to edgemaps to identify candidate transition points rapidly; second, verify these candidates using the original high-resolution frames. This segmentation allows the system to benefit from fast low-resolution processing while maintaining the precision of high-resolution analysis when needed

Inventive Principle:

Principle #1Segmentation

3Productivity

If conventional techniques use simple frame comparison methods, then the processing speed is improved, but the quality of looping video transitions deteriorates due to teleportation effects

Engineering Contradiction:

Improvelooping video generation speedVSAvoidtransition smoothness

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent introduces edgemaps as an intermediary representation between the original video frames and the transition analysis process. These edgemaps serve as a mediator that preserves edge feature information while enabling efficient comparison. By using edgemaps as the intermediary for candidate identification, the system achieves both fast processing and reliable transition detection, avoiding the teleportation effects of simple pixel-based comparison

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS10453491B2Video processing architectures which provide looping video

Publication Date: 2019.10.22 ADOBE INC

AI summary

Provided are video processing architectures and techniques configured to generate looping video. The video processing architectures and techniques automatically produce a looping video from a fixed-length video clip. Embodiments of the video processing architectures and techniques determine a lower-resolution version of the fixed-length video clip, and detect a presence of edges within image frames in the lower-resolution version. A pair of image frames having similar edges is identified as a pair of candidates for a transition point (i.e., a start frame and an end frame) at which the looping video can repeat. Using start and end frames having similar edges mitigates teleporting of objects displayed in the looping video. In some cases, teleporting during repeating is eliminated.