Julia version of cv.getPerspectiveTransform?

Specific Domains Visualization
11

Yes, ImageTransformations.jl currently just allows you to apply transformations, not fit them - this would be a nice addition to the package. Your function can be sped up with static arrays (and by rewriting the rref to use \ to solve the linear equation):

using StaticArrays

function perspective_transform_SA(
         ((x1, y1), (x2, y2), (x3, y3), (x4, y4)),
         ((X1, Y1), (X2, Y2), (X3, Y3), (X4, Y4)))

    m = SA_F64[
        x1 y1 1 0 0 0 0 0   -X1 0 0 0
        x2 y2 1 0 0 0 0 0   0 -X2 0 0
        x3 y3 1 0 0 0 0 0   0 0 -X3 0
        x4 y4 1 0 0 0 0 0   0 0 0 -X4
        0 0 0 x1 y1 1 0 0   -Y1 0 0 0
        0 0 0 x2 y2 1 0 0   0 -Y2 0 0
        0 0 0 x3 y3 1 0 0   0 0 -Y3 0
        0 0 0 x4 y4 1 0 0   0 0 0 -Y4
        0 0 0 0 0 0 x1 y1    -1 0 0 0 
        0 0 0 0 0 0 x2 y2    0 -1 0 0 
        0 0 0 0 0 0 x3 y3    0 0 -1 0 
        0 0 0 0 0 0 x4 y4    0 0 0 -1 
    ]
    v = SA_F64[0, 0, 0, 0, 0, 0, 0, 0, -1, -1, -1, -1]
    values = m \ v
    SA[
    values[1] values[2] values[3]
    values[4] values[5] values[6]
    values[7] values[8] 1
    ]
end

julia> @btime perspective_transform(((0, 0), (0, 100), (100, 0), (100, 100)),
                             ((333, 333), (345,600), (555, 346), (610, 600)))
  3.288 μs (59 allocations: 7.95 KiB)
3×3 Matrix{Float64}:
 2.50406      -0.476459    333.0
 0.307087      1.63268     333.0
 0.000511811  -0.00172887    1.0

julia> @btime perspective_transform_SA(((0, 0), (0, 100), (100, 0), (100, 100)),
                             ((333, 333), (345,600), (555, 346), (610, 600)))
  718.519 ns (0 allocations: 0 bytes)
3×3 SMatrix{3, 3, Float64, 9} with indices SOneTo(3)×SOneTo(3):
 2.50406      -0.476459    333.0
 0.307087      1.63268     333.0
 0.000511811  -0.00172887    1.0

At 718 ns per transform, that’s 1.4 million transforms per second. If you only need 400 transforms per second, this won’t be your limiting factor.

1 Like