Applies a perspective transformation to an image

dst = cv.warpPerspective(src, M)
dst = cv.warpPerspective(src, M, 'OptionName',optionValue, ...)

Input

• src Input image.
• M 3x3 transformation matrix, floating-point.

Output

• dst Output image that has the size DSize (with size(dst,3) == size(src,3)) and the same type as src.

Options

• DSize Size of the output image [w,h]. Default [0,0] means using the same size as the input [size(src,2) size(src,1)].
• Interpolation interpolation method, default 'Linear'. One of:
  • Nearest nearest neighbor interpolation
  • Linear bilinear interpolation
  • Cubic bicubic interpolation
  • Lanczos4 Lanczos interpolation over 8x8 neighborhood
• WarpInverse Logical flag to apply inverse perspective transform, meaning that M is the inverse transformation (dst -> src). default false
• BorderType Pixel extrapolation method. When 'Transparent', it means that the pixels in the destination image corresponding to the "outliers" in the source image are not modified by the function. default 'Constant'
  • Constant iiiiii|abcdefgh|iiiiiii with some specified i
  • Replicate aaaaaa|abcdefgh|hhhhhhh
  • Reflect fedcba|abcdefgh|hgfedcb
  • Reflect101 gfedcb|abcdefgh|gfedcba
  • Wrap cdefgh|abcdefgh|abcdefg
  • Transparent uvwxyz|absdefgh|ijklmno
  • Default same as 'Reflect101'
• BorderValue Value used in case of a constant border. default 0
• Dst Optional initial image for the output. If not set, it is automatically created by the function. Note that it must match the expected size DSize and the type of src, otherwise it is ignored and recreated by the function. This option is only useful when BorderType=Transparent, in which case the transformed image is drawn onto the existing Dst without extrapolating pixels. Not set by default.

The function cv.warpPerspective transforms the source image using the specified matrix:

 dst(x,y) = src((M_11*x + M_12*y + M_13) / (M_31*x + M_32*y + M_33),
                (M_21*x + M_22*y + M_23) / (M_31*x + M_32*y + M_33))

when the WarpInverse option is true. Otherwise, the transformation is first inverted with cv.invert and then put in the formula above instead of M.