| 摘要 |
The invention is to provide a method that reconstructs enhanced-resolution images with improved super-resolution algorithms, which is based on an iterative method and employs our suggested initial interpolation, robust image registration, automatic image selection, and further image enhancement post-processing. According to the method of the invention, the reconstruction of enhanced-resolution image can be quickly achieved not only by low-resolution image sequences associated with a moving object but also by low-resolution image sequences associated with a stationary scene. The invention produces an enhanced resolution image by <SL> <LI>a) storing M first low resolution images <LI>b) selecting one of the M images as a prototype image leaving a set of (M-1) second images <LI>c) interpolating extra pixels into the prototype image using the values of neighbouring pixels and calculating respective translations between the (M-1) second images and the interpolated prototype image <LI>d) dividing the translations of the (M-1) second images by a magnification factor to determine a modulus value with respect to the translations, for second images having the same modulus, selecting one of the images and grouping the selected second image and any second images not having the same modulus as N third images <LI>e) down sampling the interpolated prototype image N times to produce N fourth images <LI>f) calculating respective differences between the N third and N fourth images <LI>g) adjusting the values of pixels in the interpolated prototype image in accordance with the average of the difference values determined in step (f) <LI>h) repeating steps (e) to (g) until pixel values of the interpolated prototype image converge to a satisfactory result to yield the enhanced resolution image. </SL> |
