Research On Typical Noise Denoising For Surface Defects Image Of Strip Steel

Strip steel occupies an extremely important position in the iron and steel industry, it not only is a military, shipbuilding, aviation aerospace, machine manufacturing, chemical engineering the importance of the profession original material, but also in the car, appliance people the life essentially of the product get extensive usage. Defects on the surface of strip steel are main factors to restrict the quality improving. Its real-time monitoring, is to improve the quality of an important means. Computer vision display its importance day by day in the strip steel surface detects inspection system. Denoising for surface defects image of strip steel is surface of strip steel quality inspection system of the importance premise step, not only for the denoising filter performance requirements are relatively high, at the same time maintain its defects image detail of the information requirements of higher performance.In a detailed study of the noise model, classification, sources and strip steel surface detects inspection system in two typical noise (Gaussian noise, impulse noise), according to Gaussian noise and impulse noise with the different characteristics, targeted a different Denoising.For impulse noise, the paper on the traditional values of the Median filter and the basic principles of achieving processes, and improve its methods of identity and inadequate, and then on the basis of this comprehensive a new algorithm, a large number of simulation experiments confirmed that the method is faster computing speed, adaptive strong, denoising good effect, and maintain a good image detail features.For Gaussian noise, the paper introduces the wavelet basic theory and wavelet analysis based on wavelet denoising principle and several commonly used method. In several ways, the result of wavelet threshold denoising, a simple principle, better-performing, so through a large number of strip surface defects image simulation experiments, select better-performing small wavelet bases, decomposition level, Under different threshold and threshold functions circumstance, this method to do a detailed study.Finally, the establishment of an objective evaluation of the quality of image denoising system proved the superiority of the two denoising. The entire software system using Visual C++ and matlab achieve joint development, using Visual C++ easy to program and matlab easy to matrix calculate the characteristics, a better image denoising completed the task.