ipcv_definitions.nb

Basic image processing

Basic image display and conversion

Matrix to Mathematica Image object

$ImageObject[m_, size_:1, max_:255, min_:0] := Module[{sz, dimx, dimy, cf}, If[ ... d[sz * dimy]}, PlotRange {{0, dimx}, {0, dimy}}, AspectRatioAutomatic] ] ;$

Display Mathematica image object

Read image file to matrix

Save matrix to image file

Color to grayscale

$RowBox[{RowBox[{ColorToGrayScale[m_], :=, , RowBox[{Map, [, RowBox[{RowBox[{RowBox[{(, RowBox[{(#[[1]] + #[[2]] + #[[3]]), /, 3.}], )}], &}], ,, , m, ,, , {2}}], ]}]}], ;}]$

Grayscale to color (data conversion)

$GrayScaleToColor[m_] := Map[{#, #, #} &, m, {2}] ;$

Simple image creation

Create single color image

$SingleColorImage[rgb_, {dimx_, dimy_}] := Table[rgb, {dimy}, {dimx}] ;$

Clip single pixel/channel image values

Create Gaussian noise image

$GaussianNoiseImage[{μ_, σ_}, {dimx_, dimy_}, max_:255, min_:0] := Module[{m, ... n[μ, σ]], {dimx}, {dimy}] + offset ; Map[Clip[#, max] &, m, {2}] ] ;$

Histogram modifications

Support function for histogram plotting

$HistogramListPlot[data_, color_:Black, opts___] := Module[{g1}, g1 = Line[{{#[[1]], 0}, {#[[1]], #[[2]]}}] & /@ data ; Show[Graphics[{color, g1}], opts] ] ;$

Plot image histogram

$ShowHistogram[m_, max_:255, opts___] := Module[{mi, vals, histogram}, mi = Rou ... t[{vals, histogram} // Transpose, Blue, FrameTrue, PlotRangeAll, opts] ] ;$

Histogram equalization

$HistogramEqualization[m_, max_:255] := Module[{mi, histogram, step1, map}, mi ... #62371;map = Round[step1/Last[step1] * max] ; Map[map[[# + 1]] &, mi, {2}] ] ;$

Histogram stretch

$HistogramStretch[m_, min_:0, max_:255] := Module[{imin, imax}, imin = Min[m] ; ... x = Max[m] ; Map[((# - imin)/(imax - imin) * (max - min) + min) &, m, {2}] ] ;$

Histogram slide

$HistogramSlide[m_, offset_, max_:255] := Map[Clip[#, max] &, m + offset, {2}] ;$

Pixel modifications (apply function to individual pixels)

Gray scale: such that

Color: such that

$PixelModification[m_, f_, max_:255] := Round[Map[Clip[#, max] &, Map[f, m, {2}], {2}]] ;$

2D convolution with masks

Mask operators

Smoothing

$MeanFilter[n_] := Table[1, {n}, {n}]/(n * n) // N ;$

$GaussianFilter[n_, σ_:1] := Module[{nc, gf1}, nc = (n - 1)/2 ; gf ... /(2σ^2)], {x, 0, n - 1}, {y, 0, n - 1}] // N ; gf1/(Plus @@ Flatten[gf1]) ] ;$

Edges

Texture example

(L = Level, E = Edge, S = Spot, R = Ripple) TexureMasks: (EL, RL, ES, RR, SL, EE, ER, SR)

L5T[] := {1, 4, 6, 4, 1} ; E5T[] := {-1, -2, 0, 2, 1} ; S5T[] := {-1, 0, 2, 0, -1} ; R5T[] := ... Outer[Times, S5T[], R5T[]] ; R5S5[] := Outer[Times, R5T[], S5T[]] ; SR[] := (S5R5[] + R5S5[])/2 ;

2D convolution

Pad image with border rows so that 2D convolution does not shrink image

$PadImage[m_] := Module[{mout}, mout = Join[{First[m]}, m] ; mout = Joi ... rst[mout]}, mout] ; mout = Join[mout, {Last[mout]}] ; mout // Transpose] ;$

2D convolution of a single channel

$MaskOneChannel[mc_, kernel_] := Module[{m2}, m2 = Nest[PadImage, mc, (First[Dimensions[kernel]] - 1)/2] ; Round[Abs[ListConvolve[kernel, m2]]] ] ;$

2D convolution of an image w/kernel (image dimensions unaffected for square masks with odd-numbered dimensions)

$Mask[m_, kernel_] := Module[{r, g, b}, If[MatrixQ[m], MaskOneChannel[m ... [3]] &, m, {2}], kernel] ; Transpose /@ ({r, g, b} // Transpose) ] ] ;$

Median filtering

Filter one channel with (2n+1) x (2n+1) median filter

$MedianFilterOneChannel[mc_, n_] := Module[{ker, m1, m2}, ker = Table[0, {2 * n ... olve[ker, m1, {-1, 1}, 0, Plus, List] ; Map[Median[Flatten[#]] &, m2, {2}] ] ;$

Filter image with (2n+1) x (2n+1) median filter

$MedianFilter[m_, n_] := Module[{r, g, b}, If[MatrixQ[m], MedianFilterO ... ap[#[[3]] &, m, {2}], n] ; Transpose /@ ({r, g, b} // Transpose) ] ] ;$

Image thresholding

Simple thresholding

$ThresholdImage[m_, t_, max_:255] := Module[{m1}, m1 = If[MatrixQ[m], m, ColorToGrayScale[m]] ; Map[If[# < t, 0, max] &, m1, {2}] ] ;$

Gaussian thresholding (defined only for color images)

Mahalanobis distance from point to Gaussian distribution

Gaussian threshold an image

$GaussianThresholdImage[m_, t_, {μ_, Σ_}, max_:255] := Module[{Σi}, > ... #931;] ; Map[If[Mahalanobis[#, μ, Σi] > t, 0, max] &, m, {2}] ] ;$

Create Gaussian distance image

$GaussianDistanceImage[m_, μ_, Σ_] := Module[{Σi}, Σi = Inverse[Σ] ; Map[Mahalanobis[#, μ, Σi] &, m, {2}] ] ;$

Hough transform (straight lines)

Hough transform helper function

$ConvertEdgePixel[{xc_, yc_}, θt_, {ρmax_, dρ_}] := Module[{t1, t2}, ... (# + ρmax)/dρ] & /@ t1 ; Table[{t2[[i]], i}, {i, 1, Length[t2]}] ] ;$

From binary image, generate a sorted list of scaled line parameters

$RowBox[{RowBox[{HoughEngine[mb_, n_: {100, 100}], :=, , RowBox[{Module, [, RowBox[{{ep ... 961;}] & /@ ep, ;, , Reverse[Sort[Frequencies[Join @@ ha1]]]}]}], , ]}]}], ;}]$

From binary image, generate Hough accumulator matrix

$HoughAccumulator[mb_, n_: {100, 100}] := Module[{tmp, ha}, tmp = HoughEngine[m ... , {n[[2]]}, {n[[1]]}] ; Fold[ReplacePart[#1, #2[[1]], #2[[2]]] &, ha, tmp] ] ;$

From binary image, returns the parameters for the top , where

$RowBox[{RowBox[{StraightLines[mb_, nlines_:1, n_: {100, 100}], :=, , RowBox[{Module, [ ... 960;/nθ} & /@ Take[Last /@ HoughEngine[mb, n], {1, nlines}] //N}]}], , ]}]}], ;}]$

Visualize detected straight lines over some image

$ShowStraightLines[m_, sl_, opts___] := Module[{xdim, ydim, g1}, xdim = Dimensi ... nIdentity] ; Show[g2, g1, DisplayFunction$DisplayFunction, opts] ]$

Visualize Hough accumulator

$ShowHoughAccumulator[m_, ha_, opts___] := Module[{ρmax}, ρmax = Sqrt ... MeshRange {{0, π}, {-ρmax, ρmax}}, PlotRangeAll, opts] ] ;$

Blob finding ("Connected Component Labeling")

Identify foreground neighbors of pixel (x, y) in image

$NonZeroNeighbors[m_, {x_, y_}] := DeleteCases[{m[[y - 1]][[x - 1]], m[[y - 1]][[x]], m[[y]][[x - 1]]}, 0]$

First pass of blob finding algorithm

$BlobPass1[mc_] := Module[{d1, m2, m3, label, nzn, new}, d1 = Dimensions[mc] ; ... 2}, {d1[[1]] + 1, d1[[2]] + 1}] ; Map[If[#0, #, # - 1] &, m3, {2}] ] ;$

Unify adjacencies of labels after first pass of blob finding algorithm: support functions

$PartialAdjacency[l_] := (Union /@ (Flatten /@ Table[l[[First /@ Position[l, i]]], {i, 1, Max[l]}])) ;$

$AdjacencyRules[l_] := Table[l[[i]] Last[l], {i, 1, Length[l] - 1}] ;$

Second pass of blob finding algorithm

$BlobPass2[mc_] := Module[{tmp1, tmp2, tmp3, tmp4}, tmp1 = ListConvolve[{{0, 0} ... PartialAdjacency, Union[DeleteCases[tmp3, Null]]] ; mc /. Flatten[Union[tmp4]] ] ;$

Label all blobs with unique label in a binary image

Count the number of blobs in a labeled blob image (after BlobFind[ ])

Eliminate all but the largest blob from a labeled blob image (after BlobFind[ ])

$KeepLargestBlob[mb_] := Module[{tmp1}, tmp1 = Last /@ Drop[Reverse[Sort[Frequencies[DeleteCases[mb // Flatten, 0]]]], 1] ; mb /.((#0) & /@ tmp1) ] ;$

Display a blob image with different random color assigned to each blob

$BlobImage[mb_] := Module[{levels, colors}, levels = Drop[Sort[Union[Flatten[mb ... Integer, {50, 255}], Random[Integer, {50, 255}]}, {i, 1, Length[levels]}]] ; mb/.colors] ;$

Morphological operators for binary images

Dilation

$Dilation[mc_, kernel_, max_:255] := Module[{m2, d1, d2}, d1 = Dimensions[mc] ; ... 1 + 2 * d2] ; ListConvolve[kernel, m2, {-1, 1}, 0, BitAnd, Max] /. 1max] ;$

Erosion

$Erosion[mc_, kernel_, max_:255] := Module[{m2, d1, d2}, d1 = Dimensions[mc] ; ... elate[kernel, m2, {1, -1}, 0, BitOr[BitAnd[#1, #2], 1 - #1] &, Min] /. 1max] ;$

Closing

Opening

Created by Mathematica (February 5, 2004)