Due to the property of infrared aerial imagery, the local prior is sufficient especially for low-subrate block compressive sensing(BCS) reconstruction of infrared aerial images, while its complexity is much lower than...Due to the property of infrared aerial imagery, the local prior is sufficient especially for low-subrate block compressive sensing(BCS) reconstruction of infrared aerial images, while its complexity is much lower than nonlocal prior. The typical low-subrates can effectively improve the BCS transmission efficiency and reduce the burden of transmitter hardware. Therefore, this paper proposes a low-subrate sparse reconstruction algorithm with threshold-adaptive denoising and basis learning(TDBL), which adopts both split Bregman iteration(SBI) and adaptive threshold to implement the model-based BCS reconstruction for infrared aerial imagery. The experimental results show that as compared with the state-of-the-art algorithms, the proposed algorithm can obtain better recovery quality and less runtime on both HIT-UAV and M200-XT2 DroneVehicle datasets. the transmission efficiency and reduce the burden of transmitter hardware. The current NSS-guided reconstruction algorithms are trained and tested on natural image datasets by using relatively high subrates(more than 0.1). Due to significant difference in image contrast and pixel distribution between UAV infrared images and natural images, the performance of these algorithms on UAV infrared image datasets may be difficult to meet expectations. In recent years, the improvement of BCS recovery quality is not obvious with very high complexity, where the core step is to build a suitable dictionary, and then solve the associated sparsity of the dictionary. Previous BCS algorithms usually utilize the special iterative shrinkage/thresholding(IST)[11] method to solve the l0 minimization problem. For BCS recovery quality and runtime of UAV infrared imagery, split Bregman iteration(SBI)[12] is a competitive mechanism, so we propose the low-subrate sparse reconstruction with threshold-adaptive denoising and basis learning(TDBL) algorithm under various low-subrate cases. By analyzing the UAV infrared imagery, it is concluded that infrared aerial images are usually characterized by large number of pixels on some gray levels with double or triple peaks on the histogram, and contain more low-frequency components on the Fourier magnitude spectrum. By jointly considering both recovery quality and runtime, we solve the above l0 minimization problem of BCS reconstruction by the SBI method, instead of IST. To obtain gains during different reconstruction phases, we design an adaptive threshold ρ which is related to model-based methods, such as K-singular value decomposition(SVD) sparse coding[13] and orthogonal matching pursuit(OMP) noise constraint[14]. According to the characteristics of UAV infrared images, an updating expression of ρ is designed by combining image variance and mean value.展开更多
To progressively provide the competitive rate-distortion performance for aerial imagery,a quantized block compressive sensing(QBCS) framework is presented,which incorporates two measurement-side control parameters:mea...To progressively provide the competitive rate-distortion performance for aerial imagery,a quantized block compressive sensing(QBCS) framework is presented,which incorporates two measurement-side control parameters:measurement subrate(S) and quantization depth(D).By learning how different parameter combinations may affect the quality-bitrate characteristics of aerial images,two parameter allocation models are derived between a bitrate budget and its appropriate parameters.Based on the corresponding allocation models,a model-guided image coding method is proposed to pre-determine the appropriate(S,D) combination for acquiring an aerial image via QBCS.The data-driven experimental results show that the proposed method can achieve near-optimal quality-bitrate performance under the QBCS framework.展开更多
基金supported by the National Natural Science Foundation of China(Nos.62372100 and 62371118)。
文摘Due to the property of infrared aerial imagery, the local prior is sufficient especially for low-subrate block compressive sensing(BCS) reconstruction of infrared aerial images, while its complexity is much lower than nonlocal prior. The typical low-subrates can effectively improve the BCS transmission efficiency and reduce the burden of transmitter hardware. Therefore, this paper proposes a low-subrate sparse reconstruction algorithm with threshold-adaptive denoising and basis learning(TDBL), which adopts both split Bregman iteration(SBI) and adaptive threshold to implement the model-based BCS reconstruction for infrared aerial imagery. The experimental results show that as compared with the state-of-the-art algorithms, the proposed algorithm can obtain better recovery quality and less runtime on both HIT-UAV and M200-XT2 DroneVehicle datasets. the transmission efficiency and reduce the burden of transmitter hardware. The current NSS-guided reconstruction algorithms are trained and tested on natural image datasets by using relatively high subrates(more than 0.1). Due to significant difference in image contrast and pixel distribution between UAV infrared images and natural images, the performance of these algorithms on UAV infrared image datasets may be difficult to meet expectations. In recent years, the improvement of BCS recovery quality is not obvious with very high complexity, where the core step is to build a suitable dictionary, and then solve the associated sparsity of the dictionary. Previous BCS algorithms usually utilize the special iterative shrinkage/thresholding(IST)[11] method to solve the l0 minimization problem. For BCS recovery quality and runtime of UAV infrared imagery, split Bregman iteration(SBI)[12] is a competitive mechanism, so we propose the low-subrate sparse reconstruction with threshold-adaptive denoising and basis learning(TDBL) algorithm under various low-subrate cases. By analyzing the UAV infrared imagery, it is concluded that infrared aerial images are usually characterized by large number of pixels on some gray levels with double or triple peaks on the histogram, and contain more low-frequency components on the Fourier magnitude spectrum. By jointly considering both recovery quality and runtime, we solve the above l0 minimization problem of BCS reconstruction by the SBI method, instead of IST. To obtain gains during different reconstruction phases, we design an adaptive threshold ρ which is related to model-based methods, such as K-singular value decomposition(SVD) sparse coding[13] and orthogonal matching pursuit(OMP) noise constraint[14]. According to the characteristics of UAV infrared images, an updating expression of ρ is designed by combining image variance and mean value.
基金supported by the Natural Science Foundation of Shanghai(18ZR1400300)
文摘To progressively provide the competitive rate-distortion performance for aerial imagery,a quantized block compressive sensing(QBCS) framework is presented,which incorporates two measurement-side control parameters:measurement subrate(S) and quantization depth(D).By learning how different parameter combinations may affect the quality-bitrate characteristics of aerial images,two parameter allocation models are derived between a bitrate budget and its appropriate parameters.Based on the corresponding allocation models,a model-guided image coding method is proposed to pre-determine the appropriate(S,D) combination for acquiring an aerial image via QBCS.The data-driven experimental results show that the proposed method can achieve near-optimal quality-bitrate performance under the QBCS framework.
