From: Nicolas B. <nic...@en...> - 2005-06-08 21:19:32
|
Hi, While using vil, I've run into the following flaws : 1) vil_image_view::fill(T value) is suboptimal. Using vcl_memset when the memory is contiguous can signicantly improve the performances of this function. It can be applied very easily when sizeof(T)==1. The attached vil_fill.patch file tries to select the best implementation when the image_view is contiguous. It requires template struct within template struct support (to avoid partial specialization), I don't know is this is problematic for some compilers vxl supports. 2) I could not find any equivalent to vil1_image::in_range with vil, so I added an in_range member to vil_image_view. This is in_range.patch. 3) I could not find any vil_math_image_max/min function, the compute the max or the min between two images. Neither I could find a generic apply function, taking 2 images, a binary functor, and an output image. I think I will write one if nobody's is against it. For the moment, vil_image_math_maxmin.patch adds vil_math_image_max and min functions. I'm new to vxl and its philosophy, please let me know if I'm not respecting the spirit nor the constraints of vxl. Thanks, -- Nicolas Burrus |