研究目的
To enhance reconstruction quality in distributed compressive video sensing by proposing an adaptive reconstruction scheme that exploits temporal correlation without a feedback channel.
研究成果
The proposed adaptive reconstruction scheme based on temporal correlation improves reconstruction quality and reduces computation cost in distributed compressive video sensing, making it suitable for applications like emergency service video systems without feedback channels. It achieves performance comparable to Bi-MC algorithm with lower reconstruction time.
研究不足
The adaptive discrimination process increases computational complexity slightly, and the thresholds (T1 and T2) are set based on experimental experience, which may not be optimal for all video sequences. The scheme is primarily tested on specific sequences (Hall and Susie), and its generalizability to other videos is not fully explored.
1：Experimental Design and Method Selection:
The study uses a distributed compressed video sensing (DCVS) model with an adaptive reconstruction scheme based on temporal correlation. It employs the Gradient Projection for Sparse Reconstruction (GPSR) algorithm for reconstruction due to its acceptable quality and reduced computation cost.
2：Sample Selection and Data Sources:
The first 61 frames of the Hall sequence (288x352) and Susie sequence (240x352) are used as test sequences.
3：List of Experimental Equipment and Materials:
MATLAB R2012b software on a Windows7 (32-bit) system with Intel Core i5 processor. Wavelet transform is used for sparse transformation, and a local Hadamard matrix is used as the measurement matrix.
4：Experimental Procedures and Operational Workflow:
Key frames and non-key frames are processed using compressive sensing. Non-key frames are divided into blocks (size 16x16), classified based on temporal correlation using thresholds T1=
5：3 and T2=7, and reconstructed adaptively using GPSR. Performance is evaluated using PSNR and reconstruction time. Data Analysis Methods:
Peak signal-to-noise ratio (PSNR) and time (T) are measured and compared with existing methods (RVCS and Bi-MC algorithms).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
