作者: James F. Blinn , Michael D. Marr , Andrew C. Godfrey , Adrian Secchia
DOI:
关键词: Non-local means 、 Pixel 、 Composite image filter 、 Pixel connectivity 、 Kernel adaptive filter 、 Mathematics 、 Free boundary condition 、 Kernel (image processing) 、 Computer vision 、 Bilateral filter 、 Artificial intelligence
摘要: An optimal filter kernel, formed by convolving a box with of fixed integer width and unity area, is used to perform image resizing reconstruction. The has forced zeros at locations along frequency scale corresponding the reciprocal spacing one or more pixels that comprise source be resized. When rescale value for selected, kernel computed, mapped image, centered upon location within position an output pixel generated. number lie underneath established multiplying selected value. Upon mapping values resized are then evaluated processing pixels, such as through interpolation. Alternatively, calculated performing partial integral analysis respect standard area. each under area surrounding pixel. products summed reveal value, placed into buffer. Both these methods speed up computation process, while producing ripple free image.