Image compression sensing based on non-local feature fusion network

The emphasis of study in the area of signal processing is image compression. The traditional signal sampling has the problem of signal distortion, and the compressed sensing theory has a good application effect. To address the issue of information loss in image compression sensing construction, an approximate non-local low-rank image compression sensing model is proposed, which avoids information loss through the transformation of the similarity matrix rank minimization problem. Image reconstruction is performed by approximating the rank function with hyperbolic tangent function and compact arctangent function. Considering the insufficient performance of image compression sensing reconstruction in noisy environments, a minimization term for analyzing noise is added to the model. Considering the different distribution of noise, a regularization term is introduced to optimize the noise problem and improve the image reconstruction effect. In the performance test of image signal reconstruction model with local similarity, Monarch image is selected for reconstruction test. The proposed model has excellent image reconstruction quality, high definition of image outline and better detail retention than other models. In the image Parrotsyu reconstruction test, the proposed model has the highest reconstruction quality. When the sampling rate of SR is 0.25, the peak signal-to-noise ratio is 43.19. In the performance test of non-local image signal reconstruction model for noise analysis, the suggested model offers the best noise-cancelling performance. When the signal-to-noise ratio is 35 dB, the proposed model's peak signal-to-noise ratio is 38.65, respectively, which has the best performance among the four models. In summary, this study proposes a non-local low-rank image compression sensing model with a compact rank approximation, and adds a minimization term for analyzing noise based on it. Multiple noise fitting models are used for optimization, greatly improving the effectiveness and reconstruction quality of image compression sensing. It has important application value. The proposed compact rank approximation non local low rank image compression perception model and noise processing methods have achieved significant improvements and excellent reconstruction quality in image compression perception. These contributions are of great significance for improving the effectiveness and application value of image compression perception.