Study Of Image Adaptive Denoising Techniques Based On Dct Transform

Image is an important information carrier. In the process of acquainting andtransferring images, images are often polluted by environmental noise. How to useeffective method to eliminate the noise is a significant research work.This paper discusses the DCT adaptive de-noising. The key of DCT adaptivede-noising lies in local threshold settings, while local threshold dependent on noisevariance. Due to the noise variance estimation in DCT domain has a relatively goodestimate of the effect, whereby the DCT-1adaptive de-noising algorithm has beenproposed in this paper. Because the first DCT algorithm using the same de-noisingthreshold for homogeneous and heterogeneous region, this obviously will have overde-noising effect for heterogeneous image region, we can solve this problem as soonas smaller threshold has been accepted for heterogeneous region. It can be dividedinto two steps, the first is the distinction of image heterogeneous region, and thesecond is setting a reasonable threshold in heterogeneous image regions, and theDCT-2and DCT-3adaptive de-noising algorithms are based on this principle. TheDCT-2adaptive de-noising algorithm allows identifying heterogeneous region basedon statistical characteristics of DCT domains coefficient, it can obtain the optimalthreshold setting by doing multiple simulations. The frequency domain and spatialvariance estimation variance estimation for image heterogeneous regions havedifferent performance characteristics, so the DCT-3de-noising algorithm can identifythe image heterogeneous region based on this point, it can get the optimal thresholdparameter set after multiple sets of experiments.Because the three DCT de-noising algorithm proposed in this paper have goodrecognition of the image noise variance and heterogeneous region, compared with thetraditional blind de-noising algorithms, these three adaptive de-noising algorithmsproposed in this paper have a much better de-noising effects. The latter two de-noisingalgorithm have done processing for image heterogeneous region, so they are betterthan the first de-noising noising algorithm. For most of the image, the DCT-3hasmore accuracy of image heterogeneous regions, so its de-noising effect is better than rthe DCT-2’s.