| 摘要 |
Method for analysing an image by estimating a continuous function, centred on a selected pixel, of a pixel attribute (eg. chrominance, luminance), comprising: determining nested pixel sets with a shared centre at a common pixel; estimating a discrete weighting for each set depending on the function; mapping the sets on to pixels of an image, with the common pixel mapped to the selected pixel position; combining the pixel attribute values for each set; summing the weighted combined values to approximate the function. Concentric pixel sets may be annular or ring shaped. Weighting may be calculated and stored prior to acquiring the picture. Weighting for each pixel set may equal the function value at the sets mid-point. The equation may be an isotropic radially symmetric function around the central pixel. Pixels in the sets may be at varying distances from the central picture element. Visual boundaries or edges may be detected in an image. The formula may be a filter such as Gaussian, Laplacian of Gaussian, difference of Gaussians, determinant of Hessian, Sobel, Canny and Canny-Deriche. Regions within detected borders may be coloured with an average hue. Neighbouring pixels may be added, subtracted or deducted when subsequent pixels are analysed. |
