| 摘要 |
An image manipulation process, such as interpolation of an image in a sequence, depends on a spatial periodicity measure for image data that is a function of the sparseness of the two-dimensional spatial frequency spectrum of the processed image data. At least one two-dimensional (2D) block of image data may be processed in a two-dimensional Fourier or other spatial to frequency domain transform and preferably the periodicity measure is evaluated from the portion of the transformed image data representing AC spectrum components. Sparseness may be a linear function of the mean-square value and the square of the mean value of transformed image data for the 2D block; sparseness can further depend on the mean-square value of low spatial frequency transform output values. Sparseness can be a count of the number of spatial frequency values exceeding a threshold, and may be measured by allocating spectrum values to frequency bins, counting the number of non-zero values. The spatial periodicity measure may be normalised, divided by a function of the mean-squared image data. A two-dimensional spatial frequency spectrum having all equal values has zero sparseness; a spectrum having only one non-zero value has a unity sparseness. |
