论文部分内容阅读
针对压缩遥感过程中非严格稀疏和傅里叶域欠采样噪声导致的伪影和混叠现象,提出了基于梯度转向核的压缩重构策略(GradSK).在压缩感知编码过程中提出了半随机傅里叶测量的方式,既保留图像的概要分量,同时保证了K-空间随机欠采样的非连贯性.在压缩感知解码过程中提出了由基于多阶梯度的转向核与有限差分总方差(TV)结合的方法,来解决解码过程中的无约束凸框架问题.实验表明,该方法在解决无噪采样和有噪采样的过程中均有较好性能.
Aiming at the artifacts and aliasing caused by non-strict sparsity and under-sampled noise in Fourier domain for compressive remote sensing, a compressive reconstruction strategy (GradSK) based on gradient steering kernel is proposed. In the process of compressive sensing coding, semi-random The Fourier transform method not only preserves the summary components of the image but also ensures the non-coherent random under-sampling in K-space.In the process of compressed sensing decoding, a new method based on multi-step gradient steering kernel and finite difference total variance TV) to solve the unconstrained convex frame problem in decoding process.Experiments show that this method has better performance in solving both no-noise and noisy samples.