Abstract:
The present invention is directed to an image processing method, comprising: deriving a pixel statistical value of pixels and edge information for each of regions in a plurality of layers, the regions including attention pixels and having ranges that are successively narrower; correcting differential information between a pixel statistical value for a region in an attention layer and a pixel statistical value for a region in a layer wider than the region in the attention layer by using the edge information; correcting the pixel statistical value for the region in the attention layer by using the corrected differential information and the pixel statistical value for the region wider than the region in the attention layer; re-correcting the corrected pixel statistical value for the region in the attention layer by using a pixel statistical value for a region equal to or wider than a region in each of the layers and differential information between the uncorrected pixel statistical value for the region in the attention layer and the corrected pixel statistical value for the region in the attention layer; and correcting the attention pixel by repeating the correcting and the re-correcting the pixel statistical value for the region in the attention layer sequentially in the respective layers until the region is reduced from a maximum range to a minimum range.
Abstract:
A method includes: calculating a pixel statistical value and edge of pixels for each of areas of a multi-layer, the areas each containing a target pixel and having a successively decreased range; correcting the edge based on a pixel statistical value of an area that is wider than an area of a specific layer; correcting difference between a pixel statistical value of the specific layer and the pixel statistical value of a layer that is wider than the specific layer using the post-correction edge; correcting the pixel statistical value of the specific layer using post-correction difference and the pixel statistical value of the layer that is wider than the specific layer; and correcting the target pixel by repeating correction of the pixel statistical value successively in each layer until the area reduces its range from the maximum range to the minimum range.
Abstract:
Disclosed is an image processing method including: generating an initial denoised image with a reduced noise while preserving an edge in an input image; controlling an iterative operation performed based on energy defined in advance based on an initial residual component calculated from the input image and the initial denoised image; and separating the initial denoised image to a skeleton component and a residual component by the controlled iterative operation to generate the skeleton component as an output image.