Method of adaptive noise smoothing/restoration in spatio-temporal domain and high-definition image capturing device thereof

2001 
PURPOSE: A spatiotemporal adaptive filtering/high density recovering method and a high density image input device by the same are provided to effectively remove the color stains and poison noise while preserving the edge lines and detail information for recovering images as captured under bright illumination. CONSTITUTION: A spatiotemporal adaptive filtering/high density recovering method includes the steps of sensing the degree of a motion by computing an intensity difference and a chromaticity difference between respective pixels which compose a comparing frame and respective pixels of a reference frame of an image, calculating a chromaticity weighting function and a chromaticity weighting function in intensity weighting function and chromaticity weighting function operation blocks(110,130) according to the motion degree sensed by the differences between the reference and comparing frames, carrying out temporal filtering for pixels of which motion degrees are lower than selected threshold values for the selected number of frames with respect to R,G,B channels, converting the image from RGB into YUV format in a YUV conversion part(200), sensing an edge degree by computing an intensity difference between the predetermined number of selected environmental pixels and pixels forming a predetermined frame of the image, calculating an intensity weight function according to the sensed edge degree, calculating a local mean or a local variance in local mean and variance calculating parts(400,500), carrying out LLMMSE filtering for the intensity components of the image by the local mean or the local variance, and converting the intensity components into the R,G,B format in an RGB conversion part(600) by adding chromaticity components before the filtering.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    0
    References
    0
    Citations
    NaN
    KQI
    []