Shape Contour with redCV

Another way to analyse shapes in image is to calculate the signature of the shape. Basically, the centroid of the shape given by the rcvGetCentroid is the origin of polar coordinates system. Consequently all pixels that belong to the contour of the shape can be described by the distance rho to the origin according to the angle theta.

RedCV proposes 2 fonctions for that:

rcvGetEuclidianDistance, which get the rho distance of the pixel (as-pair) to the centroid of the shape.

rcvGetAngle which calculate the theta angle (0..359 °) of the pixel.

© Bruno Keymolen 

As demonstrated here, redCV makes the job correctly. When looking for the signature of a circle, we get a line since all pixels are equidistant from the centroid.

For a square, it is also correct and the four angles of the square are clearly identifiable.

Code sample

Red [
    Title:   "Matrix tests "
    Author:  "Francois Jouen"
    File:    %freeman2.red
    Needs:   'View
]


#include %../../libs/redcv.red ; for redCV functions

iSize: 512x512
mat:  rcvCreateMat 'integer! 32 iSize
bMat: rcvCreateMat 'integer! 32 iSize
img: rcvCreateImage iSize
plot:  copy [fill-pen white box 155x155 355x355]
_plot: copy [line-width 1 pen green 
            text 175x480 "Angle"
            line 5x10 5x470 5x470 375x470 375x5 5x10 
            line 190x10 190x470
            text 10x450 "0" text 178x450 "180" text 345x450 "360" 
            line]
plot2: copy _plot
fgVal: 1
canvas: none


processImage: does [
    img: to-image canvas
    rcvImage2Mat img mat     
    rcvMakeBinaryMat mat bmat
    cg: rcvGetMatCentroid bmat img/size     ; get shape centroid
    border: []
    rcvMatGetBorder bmat iSize fgVal border ; get border
    angles: copy []
    foreach p border [
        ; use x y coordinates and calculate rho and theta
        rho: rcvGetEuclidianDistance p cg
        theta: rcvGetAngle p cg
        bloc: copy []
        append bloc theta
        append bloc rho
        append/only angles bloc 
    ]
    sort angles ; 0.. 359  to use with line draw command
    foreach n angles [
        p: as-pair first n 384 - second n 
        p: p + 10x0
        append plot2 (p)
    ]
    canvas2/draw: reduce [plot2]
]



; ***************** Test Program ****************************
view win: layout [
    title "Contour Signature"
    
    r1: radio "Square" [canvas/image: none 
                        canvas2/image: none
                        plot: compose [fill-pen white box 155x155 355x355]
                        plot2: copy _plot
                        canvas/draw: reduce [plot]
                        canvas2/draw: reduce [plot2]
                        ]
    r2: radio "Circle" [canvas/image: none 
                        canvas2/image: none
                        plot: compose [fill-pen white circle 255x255 120] 
                        plot2: copy _plot
                        canvas/draw: reduce [plot]
                        canvas2/draw: reduce [plot2]
                        ]
    r3: radio "Triangle" [canvas/image: none 
                        canvas2/image: none
                        plot: compose [pen white fill-pen white triangle 256x128 128x300 384x400] 
                        plot2: copy _plot
                        canvas/draw: reduce [plot]
                        canvas2/draw: reduce [plot2]
                        ]
    r4: radio "Polygon" [canvas/image: none 
                        canvas2/image: none
                        plot: compose [pen white fill-pen white polygon 256x100 384x300 128x400 128x300 256x10] 
                        plot2: copy _plot
                        canvas/draw: reduce [plot]
                        canvas2/draw: reduce [plot2]
                        ]
    button "Process" [processImage]
    pad 395x0
    button "Quit" [ rcvReleaseImage img
                    rcvReleaseMat mat
                    rcvReleaseMat bmat
                    Quit]
    return
    canvas: base 512x512 black draw plot
    canvas2: base 380x512 black draw plot2
    do  [r1/data: true]
]