Gaussian noise, salt and pepper noise, and the signal sparsity problem exist in the strip surface image . Therefore, a de-noising method using compressed sensing algorithm is researched in this paper . An image de-noising model is established based on Stagewise Regularized Orthogonal Matching Pursuit algorithm. The simulation experiments and comparative analysis for three typical defect images such as edge cracks, holes and roll mark are carried out. Experiment results show that the signal can still reconstructed effectively and reliably when the signal sparsity is unknown. The presented algorithm can improve the computation speed and ensure global optimization. Especially, the peak signal to noise ratio (PSNR) values is higher. Then, the presented algorithm can effectively filter out the noise pollution, improve the image quality, and can meet the requirement of real-time image processing.