METHOD AND SYSTEM FOR GENERATING HIGH RESOLUTION PEAK ACTION FRAME

Smart overview of the Invention

The patent application describes a method and system for generating high-resolution peak action frames. This involves processing both low-resolution (LR) and high-resolution (HR) image streams, where the LR stream has a high frame rate and the HR stream has a lower frame rate. The system detects human poses in the LR stream to identify peak action frames, then finds the closest corresponding HR frame. The HR peak action frame is generated by either selecting the closest HR frame or blending it with the LR peak action frame.

Background

The need for high-resolution images of peak action moments, such as the highest point of a jump, is common in various applications. Current systems can capture LR frames at high speeds but struggle with capturing HR frames at similar rates. Thus, HR peak action frames are often missed. The proposed method aims to overcome these limitations by effectively combining LR and HR streams to capture these critical moments.

Technical Approach

The method involves several key steps: receiving LR and HR image streams, detecting human poses in the LR stream, and identifying LR peak action frames. An AI module tracks the action's path, predicting the time and location of peak actions. The closest HR frame is identified, and a HR peak action frame is generated by blending this frame with the LR peak action frame. This approach ensures high-resolution capture of peak actions.

Additional Features

Beyond peak action frame generation, the system can predict slow-motion start and end times based on peak action frame arrival. It also adjusts frame rates to create slow-motion video segments. The method uses deep neural networks to analyze poses and action cues, enhancing the accuracy of peak action detection. This comprehensive approach improves the quality and utility of video captures.

Device Implementation

An electronic device implementing this method includes memory and processors configured to execute the described instructions. The device receives LR and HR streams, detects poses, and identifies peak action frames. It adjusts zoom factors based on action cues and modifies frame rates for slow-motion segments. This integration of hardware and software capabilities enables efficient high-resolution peak action frame generation.