man pages section 3: Multimedia Library Functions

Exit Print View

Updated: July 2014



mlib_ImageGridWarp_Fp - grid-based image warp


cc [ flag... ] file... –lmlib [ library... ]
#include <mlib.h>

mlib_status mlib_ImageGridWarp_Fp(mlib_image *dst, const mlib_image *src,
     const mlib_f32 *xWarpPos, const mlib_f32 *yWarpPos,
     mlib_d64 postShiftX, mlib_d64 postShiftY,
     mlib_s32 xStart, mlib_s32 xStep, mlib_s32 xNumCells,
     mlib_s32 yStart, mlib_s32 yStep, mlib_s32 yNumCells,
     mlib_filter filter, mlib_edge edge);


The mlib_ImageGridWarp_Fp() function performs a regular grid-based image warp on a floating-point image. The images must have the same type, and the same number of channels. The images can have 1, 2, 3, or 4 channels. The data type of the images can be MLIB_FLOAT or MLIB_DOUBLE. The two images may have different sizes.

The image pixels are assumed to be centered at .5 coordinate points. For example, the upper-left corner pixel of an image is located at (0.5, 0.5).

For each pixel in the destination image, its center point D is, first, backward mapped to a point S in the source image; then the source pixels with their centers surrounding point S are selected to do one of the interpolations specified by the filter parameter to generate the pixel value for point D.

The mapping from destination pixels to source positions is described by bilinear interpolation between a rectilinear grid of points with known mappings.

Given a destination pixel coordinate (x, y) that lies within a cell having corners at (x0, y0), (x1, y0), (x0, y1) and (x1, y1), with source coordinates defined at each respective corner equal to (sx0, sy0), (sx1, sy1), (sx2, sy2) and (sx3, sy3), the source position (sx, sy) that maps onto (x, y) is given by the formulas:

  xfrac = (x - x0)/(x1 - x0)
  yfrac = (y - y0)/(y1 - y0)

  s = sx0 + (sx1 - sx0)*xfrac
  t = sy0 + (sy1 - sy0)*xfrac

  u = sx2 + (sx3 - sx2)*xfrac
  v = sy2 + (sy3 - sy2)*xfrac

  sx = s + (u - s)*yfrac - postShiftX
  sy = t + (v - t)*yfrac - postShiftY

In other words, the source x and y values are interpolated horizontally along the top and bottom edges of the grid cell, and the results are interpolated vertically:

         (x0, y0) ->            (x1, y0) ->
           (sx0, sy0)             (sx1, sy1)
            |           /|         |
            |     (s, t) |         |
            |            |         |
            |            |         |
            |            |         |
            |            |         |
            | (x, y) ->  |         |
            |  (sx, sy)--+         |
            |            |         |
            |            |         |
            |            | (u, v)  |
            |            |/        |
         (x0, y1) ->          (x1, y1) ->
           (sx2, sy2)           (sx3, sy3)

The results of above interpolation are shifted by (-postShiftX, -postShiftY) to produce the source pixel coordinates.

The destination pixels that lie outside of any grid cells are kept intact. The grid is defined by a set of equal-sized cells. The grid starts at (xStart, yStart). Each cell has width equal to xStep and height equal to yStep, and there are xNumCells cells horizontally and yNumCells cells vertically.

The degree of warping within each cell is defined by the values in xWarpPos and yWarpPos parameters. Each of these parameters must contain (xNumCells + 1)*(yNumCells + 1) values, which, respectively, contain the source X and source Y coordinates that map to the upper-left corner of each cell in the destination image. The cells are enumerated in row-major order. That is, all the grid points along a row are enumerated first, then the grid points for the next row are enumerated, and so on.

For example, suppose xNumCells is equal to 2 and yNumCells is equal to 1. Then the order of the data in the xWarpPos would be:

  x00, x10, x20, x01, x11, x21

and in the yWarpPos:

  y00, y10, y20, y01, y11, y21

for a total of (2 + 1)*(1 + 1) = 6 elements in each table.


The function takes the following arguments:


Pointer to destination image.


Pointer to source image.


A float array of length (xNumCells + 1)*(yNumCells + 1) containing horizontal warp positions at the grid points, in row-major order.


A float array of length (xNumCells + 1)*(yNumCells + 1) containing vertical warp positions at the grid points, in row-major order.


The displacement to apply to source X positions.


The displacement to apply to source Y positions.


The minimum X coordinate of the grid.


The horizontal spacing between grid cells.


The number of grid cell columns.


The minimum Y coordinate of the grid.


The vertical spacing between grid cells.


The number of grid cell rows.


Type of resampling filter. It can be one of the following:


Type of edge condition. It can be one of the following:


Return Values

The function returns MLIB_SUCCESS if successful. Otherwise it returns MLIB_FAILURE.


See attributes(5) for descriptions of the following attributes:

Interface Stability

See also

mlib_ImageGridWarp(3MLIB), mlib_ImageGridWarpTable(3MLIB), mlib_ImageGridWarpTable_Fp(3MLIB), attributes(5)