Compressive Sensing Image Denoising Algorithm And Its Application Over Wiressless Fading Channels

Channel is the signal transmission channel in communication system, which is composedof the transmission medium that signals go through from the transmitter to the receiver.Because channels usually exist noise, the most common channel noises are gaussian noise andimpulse noise, or mixed noise, that makes signal in the transmission have noise. This papermainly discuss digital image, so, noise signal is noise image, noise image seriously affectedpeople’s visual effect, hindered the recognition of people, in order to improve the quality ofthe image so that further processing to the image can be made, restraining channel noise is anecessary part.The paper first introduces the traditional image denoising algorithms, mainly discussedaverage filtering, median filtering and wiener filtering, Fourier denoising transform andwavelet denoising algorithm. average filter and wiener filtering can suppress gaussian noisewell, but is only suitable for low-intensity gaussian noise, and has bad anti-noise ability toimpulse noise, median filter can control the salt and pepper noise well, if the noise intensity issmall, it can also keep the details and edge of image. but is not suitable for removing gaussiannoise and high-intensity salt and pepper noise. the Fourier transform is a globaltransformation, unable to express time-frequency local properties of signal, while, wavelettransform not only has the characteristics of multi-resolution analysis, in both time andfrequency, but also can express time-frequency local properties of signal, so the waveletdenoising method compared with the previously mentioned methods, can better preserve theimage edges and details, but when the noise intensity increased to a certain extent, the waveletdenoising method is not applicable, because it will cause serious loss of information, and thedenoising effect is also not good.Aiming at the shortcomings of the mentioned above, the paper proposed improved BasisPursuit denosing(IBPDN), which changes L2form to noise to L1form, IBPDN is areconstruction algorithm of compressive sensing. The experimental data show that IBPDNcan control high-intensity impulse noise and gaussian nosie in the channels, furthermore,more satisfactory results are obtained using IBPDN to mixed noise than those with BPDN.and also keep edges and details of image.