Video Compressed Sensing Reconstruction via Frame Fragment Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compressed sensing algorithms are sensitive to computing complexity and suffer from slow rendering and reconstruction speeds, even with graphics processing units, and often result in low reconstruction quality.
Innovation Solution
A method and system that extracts frame fragments from compressed video frames using a predetermined rule, performs feature abstraction through multiple hidden layers of a pre-trained video frame reconstruction model, and reconstructs these fragments into blocks, reducing computing complexity and improving quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If video compressed sensing algorithms process entire video frames, then reconstruction quality can be maintained, but computing complexity increases and reconstruction speed decreases
Solution Approach 1:
The patent divides video frames into multiple frame fragments and processes them separately through the reconstruction model. This segmentation approach reduces the computational burden on each processing step while maintaining overall reconstruction quality, as the model can focus on smaller, more manageable portions of the video data at any given time.
Solution Approach 2:
The patent employs a pre-trained video frame reconstruction model that has been prepared in advance through training on video data. This preliminary action of pre-training allows the model to learn effective reconstruction patterns beforehand, enabling faster and more efficient real-time reconstruction without requiring complex computations during the actual reconstruction process.
2Manufacturing precision
If video compressed sensing algorithms process entire video frames, then complete video information is reconstructed, but reconstruction speed becomes extremely slow
Solution Approach 1:
The patent segments video frames into multiple smaller frame fragments that are processed independently and in parallel. This segmentation enables the reconstruction system to handle multiple fragments simultaneously, significantly improving reconstruction speed while maintaining the quality of each fragment through the pre-trained model.
Solution Approach 2:
The pre-trained reconstruction model serves itself by automatically learning optimal reconstruction patterns during the training phase. Once trained, the model can independently process frame fragments without requiring complex real-time adjustments or additional computational resources, thereby achieving fast and efficient reconstruction.
3Productivity
If graphics processing units are used for parallel acceleration, then computing speed improves, but reconstruction quality remains relatively low
Solution Approach 1:
The patent performs the computationally intensive training process in advance to pre-train the reconstruction model. This preliminary action transfers the computational burden from the real-time reconstruction phase to the offline training phase, allowing fast parallel processing during actual use while maintaining high reconstruction quality through the learned model parameters.
Solution Approach 2:
The patent replaces traditional mechanical parallel processing approaches with an intelligent neural network model that has learned reconstruction patterns. This substitution allows the system to achieve both high speed (through efficient model inference) and high quality (through learned optimal reconstruction strategies) without relying solely on brute-force parallel computation.
Data Source
AI summary
The present disclosure provides a video compressed sensing reconstruction method, including: step B, after receiving to-be-reconstructed compressed video frames, extracting frame fragments of the compressed video frames according to a predetermined extraction rule; step C, inputting the frame fragments into an input layer of a pre-trained video frame reconstruction model, performing feature abstraction to the frame fragments through multiple hidden layers of the video frame reconstruction model, and building a nonlinear mapping between each frame fragment and a corresponding frame fragment block; and step D, reconstructing the input frame fragments to frame fragment blocks by the hidden layers according to the nonlinear mapping, and outputting the frame fragment blocks by an output layer of the video frame reconstruction model, and generating a reconstructed video based on the reconstructed frame fragment blocks. The present disclosure can render and reconstruct video frames quickly with a high quality.