Research On Compressed Sensing Theory And Its Application In Image De-noising

With the extension of service time of highway，the pavement affected by various factorswill appear disease phenomenon，such as crack, etc. Timely and effective for pavementdisease detection, can greatly reduce the disease worsens and ensure driving safety. However,due to the influence of all sorts of noises, the crack image is having characteristics such asdiversity and complexity, which brings difficulty into researching efficient pavementdetection algorithm. Therefore, how to eliminate the noise of the image has become the hotissues in looking into pavement crack detection. In this paper, the research on compressedsensing theory and its application in pavement crack image de-noising will have greatsignificance in practical detection.Compressed Sensing is a new set of theory in signal sampling. It breaks through thebottleneck of the Nyquist sampling theorem and puts forward that if the measurement matrixis far less than the Nyquist sampling frequency in some certain conditions, then we are able torecover the original signal accurately by making use of the discrete samples which get fromrandom sampling of signal and through non-linear reconstruction algorithm. As a result, inorder to overcome the drawbacks of conventional algorithms, this paper applies thecompressed sensing theory to the research problem on image de-noising.Based on the review of current research status of image de-noising both in domestic andabroad, first of all, this paper introduces the image de-noising methods in spatial domain andtransform domain, in which focuses on describing wavelet threshold, then compares andanalyzes the algorithm performance. Secondly, the paper in detailed explains the compressedsensing theory, which core is signal sparse representation, observation matrix design andrelated algorithm of signal reconstruction. Finally, the paper puts forward a new pavementcrack image de-noising method based on the compressed sensing theory. Under the frame, itfocuses mostly on the research of gradient projection for sparse reconstruction algorithm.Furthermore, it also analyses an image de-noising method on the basis of TV norm to improvealgorithm. The result of the simulation tests verify the effectiveness and feasibility of thealgorithm based on compressed sensing theory in image de-noising. In addition, it also teststhat the gradient projection reconstruction algorithm based on TV norm, in which can betterreserve the effective information meanwhile removing noise, is more suitable forde-noising processing on pavement crack image.