In radar system, the maneuvering targets undergoing significant angular motion pose a challenge for motion compensation, thus an inverse synthetic aperture radar (ISAR) imaging framework has been proposed to obtain focused ISAR images under sparse aperture conditions. Due to the non-stationary Doppler frequency induced by the targets' complex motion, conventional range Doppler (RD) algorithm suffers from performance degradation. Therefore, we propose a suitable coherent processing interval (CPI) extraction method to find out the relatively stable Doppler range. Since the fact that the extracted CPI typically contains too limited pulses to generate high-quality ISAR images, a block-sparse signal recovery method based on pattern-coupled sparse Bayesian learning strategy for matrix block (PC-MBSBL) is further developed, which introduces the improved pattern-coupled hierarchical Gaussian model as the sparse prior to reconstruct the scattering coefficient. Specifically, the sparsity of each coefficient is represented by the hyperparameters corresponding to its eight adjacent coefficients, rather than solely by its own hyperparameter. The proposed prior model provides flexibility in modeling any block-sparse signals, and facilitates the generation of block structures. The experimental results using simulated data and measured data demonstrate the validity and superiority of the proposed method.

Image acquired by Camera sensors suffers from low contrast and poor visibility under low-light conditions. Given the importance of enhancing low-light images, this paper presents an innovative method known as the self-calibrating neural network (SCNet). This approach involves the extraction of a lighting adjustment feature map from the input image. Matrix operations are then employed on the input image to derive an intermediate image, which is subsequently optimized through self-calibration to produce the final output image. The SCNet approach has proven effective in experimentation, as indicated by the peak signal-to-noise ratio (PSNR) values, demonstrating a 6.1% increase compared with the baseline on the low-light (LOL) dataset. The SCNet can be trained without the necessity for reference images, overcoming the challenge of insufficient enhancement often observed in very low-light images, which is a prominent problem in existing unsupervised low-light enhancement methods.