$L_0$ Sparse Regularization-Based Image Blind Deblurring Approach for Solid Waste Image Restoration

The new vision-based object sorting system is a fundamental module in the construction and demolition waste recycling industry, where image deblurring is vital as the system often fails due to the heavily blurred images caused by the vibration of the conveyor belt that carries solid wastes. This paper proposes a novel blind deblurring approach in which a novel penalty function is formulated as the regularization term in the total energy function. This regularization term is based on sparse prior and solved as part of a mathematical optimization problem, which is operated on the dark channel of the input image. The method not only reserves the structure information of the image but also avoids over-smoothing in the final restoration. On synthetic and natural blurred images, the method outperforms other popular methods. Even with less iterations, the convergence rate and quality of the results are superior. We apply this approach for solid waste image restoration and achieve remarkable results with high validity and reliability.