Clutter And Artifact Reduction Research Of Medical CT Image Based On Singular Value Filter

Nowadays, medical CT imaging technology has become the most prevalent assistance tool forclinical diagnosis. But clutters and artifacts are inevitably introduced to medical CT images inmedical imaging process, which results from imaging system and external interference factors andso on. Clutter artifacts in medical CT images severely degrade pixel density resolution of image,which results in image blurred in lesion regions of physiological structure image, bringing doctortrouble that they can’t fast diagnosis diseases.Based on the characteristics of medical CT image, thus the paper proposes a singular valuefiltering framework method, which can effectively reduce artifact of CT image and can also removenoise of CT image. This paper works as follows:Firstly, the paper pointed out a filting method based on filtered backprojection and singularvalue decomposition by blocking image to reject star artifact in the medical CT image. The processof backprojection reconstructing image usually results in star artifact.Thus, above filteredbackprojection reconstructing image thesis, the paper uses singular value decomposition to imageblock, which makes a good effect in rejecting star artifact.Secondly, it was used an algorithm based on nearest neighbor interpolation and singular valuedecomposition to reduce ring artifact in CT image. Image was reconstructed by back-projection tothe original shepp-logan phantom image in order to image with ring artifact, which used robertsedge detection operator function for image artifacts edge and mask them. Nearest neighbor wasinterpolated to ring artifacts and finally it was used block singular value decomposition filtering forreconstructing image.Thirdly, the thesis put forward an improvement method about reconstruction order for imagedenoising by singular value decomposition. The traditional methods on determining number ofeffective singular value was according to the curve inflection points of singular value spectrum andsingular entropy increment as threshold value. But sometimes the inflection points were not clear,i.e., taking on “arc transition zone”, especially in the image with high peak SNR. In order to makeup the deficiencies, it was proposed a method to determine threshold by the extreme points ofreconstructing image using singular value decomposition.Fourthly，it was pointed out an denoising algorithm based on improved singular value filterfunction in CT image. The traditional singular value filters usually retained singular values above the threshold and rejected singular values below the threshold. According to improveed signalmodel of noise in CT image, the paper proposed a new function for singular value filter, whichcould effectively remove the noise and also retain the useful information of the image as far aspossible.