研究目的
To present a novel drone high dynamic range infrared video coding algorithm based on aerial map prediction that improves coding performance by utilizing global motion estimation and extra reference frames from an aerial map.
研究成果
The proposed aerial map prediction algorithm provides significant improvements in coding efficiency for drone high dynamic range infrared videos, with 3–35% bitrate savings compared to H.265/HEVC for videos with camera rotation. Future work includes enhancing the aerial map building algorithm and applying more sophisticated motion models.
研究不足
The coding performance significantly depends on the quantization parameter for the aerial map, requiring joint optimization for best performance. The approach may not provide advantages if the aerial map bitrate is higher than the bitrate savings provided by the AMP.
1：Experimental Design and Method Selection:
The proposed algorithm accumulates input frames in a buffer to build an aerial map, applies global motion estimation to extract similar frames for use as extra reference frames in H.265/HEVC coding.
2：Sample Selection and Data Sources:
Two test video sequences captured by a Flir Tau2 infrared camera with frame resolution 640×512, 16 bits per pixel.
3：List of Experimental Equipment and Materials:
Flir Tau2 infrared camera, Drone Systems ApS drone.
4：Experimental Procedures and Operational Workflow:
Frames are aligned using rotation angle, vertical and horizontal displacements; aerial map pixels are computed as a weighting combination of corresponding pixels in overlapped frames.
5：Data Analysis Methods:
Rate-distortion performance evaluated using Peak Signal-to-Noise Ratio (PSNR) as an objective quality metric.
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
