Image processing method and imaging apparatus
An image processing method for reducing noise of an image to be filtered; the method includes a step of creating shrunken image of shrinking ratio r×r (where r is a natural number) by averaging the pixel values for each pixel in a block for every r pixels, a step of obtaining the pixel value in the shrunken image that correspond to the focus pixel in the source image by performing linear interpolation from the shrunken image pixels, a step of determining the noise reduction filter from the focus pixel and reference pixel in the shrunken image, and a step of computing the compensated pixel value of the focus pixel in the source image by weighted sum of each pixel value applying the filter to the focus pixel in the source image and the corresponding shrunken pixel.
1. An image processing method for reducing noise of an image to be filtered by determining a noise reduction filter coefficient and calculating weighted sum, with the coefficient, of the pixel values of a focus pixel and its surrounding reference pixels of the image, comprising:
a step of determining a first coefficient of the noise reduction filter by using the difference in pixel values and relative position between a focus pixel and each of its surrounding reference pixels in the source image where the source image is the target image to be filtered;
a step of creating a smoothed image by smoothing pixel values within a surrounding area for each pixel in the source image;
a step of determining a second coefficient of the noise reduction filter by using the difference in pixel values and relative position between a focus pixel of the source image and a pixel extracted, by every predetermined number of pixels, from the smoothed image; and
a step of computing the compensated pixel value of the focus pixel in the source image with weighted sum of each pixel value applying the first coefficient to the reference pixels and the second coefficient to the pixels detected from the smoothed image.
2. An image processing method for reducing noise of an image to be filtered by determining a noise reduction filter coefficient and calculating weighted sum, with the coefficient, of the pixel values of a focus pixel and its surrounding reference pixels of the image, comprising:
a step of determining a first coefficient of the noise reduction filter by using the difference in pixel values and relative position between a focus pixel and each of its surrounding reference pixels in the source image where the source image is the target image to be filtered;
a step of creating a shrunken image by smoothing pixel values for every block of a predetermined size in the source image;
a step of obtaining the pixel value in the shrunken image that correspond to the source image pixel by performing linear interpolation from the shrunken image pixels;
a step of determining the second coefficient of the noise reduction filter by using the difference in pixel values and relative position between a focus pixel of the source image and a pixel of a certain coordinate related to the focus pixel in the shrunk image; and
a step of computing the compensated pixel value of the focus pixel in the source image by weighted sum of each pixel value applying the first coefficient to the focus pixel and the second coefficient to the pixel of the predetermined coordinate in the linear-interpolated shrunken image.
3. The image processing method according to claim 1 or 2 , further comprising:
a step of detecting the direction of an edge in the focus pixel by applying an edge detection filter pre-stored; and
a step of determining the noise reduction coefficient from the relative position between the focus pixel and each reference pixel according to the detected edge direction.
4. The image processing method according to claim 3 :
wherein the method stores a distance table that indicates the relative position between a focus pixel and each reference pixel for every edge direction, and references a distance table for the appropriate detected edge direction to determine said relative position.
5. An image processing apparatus to create a compensated image by reducing noise in a captured image according to the difference in pixel values and relative position between a focus pixel and each of its surrounding reference pixels in the captured image, comprising:
a means for storing edge detection filters;
a means for storing a distance table that indicates the relative position between a focus pixel and each reference pixel for every edge direction;
an image shrinking means for creating a shrunken image by smoothing pixel values for every block of a predetermined size in the source image where the source image is the target image to be filtered;
an edge direction detection means for detecting the edge direction and strength of the focus pixel by using the edge detection filter;
a position weight filter composite means for creating a position weight filter from the detected edge strength and the relative position determined by the distance table corresponding to the edge direction detected; and
a pixel value calculation means for computing the pixel value in the shrunken image that correspond to the source image pixel by performing linear interpolation from the shrunken image pixels, and computing the compensated pixel value of the focus pixel in the source image according to the pixel values obtained by applying the position weight filter to a reference pixel of the source image and the pixel values obtained by applying the position weight filter to a pixel of a certain coordinate related to the focus pixel in the shrunk image.