[ImgProc] some functions should be in the white list

[ganwenyao]

Function Name	Location
cv.getRotationMatrix2D()	Geometric Transformations
cv.getPerspectiveTransform()	Geometric Transformations
cv.filter2D()	Smoothing Images
cv.morphologyEx()	Morphological Transformations
cv.getStructuringElement()	Morphological Transformations
cv.convertScaleAbs()	Image Gradients
cv.isContourConvex()	Contours
cv.minAreaRect()	Contours
cv.minEnclosingCircle()	Contours
cv.fitLine()	Contours
cv.minMaxLoc()	Contours
cv.convexityDefects()	Contours
cv.pointPolygonTest()	Contours
cv.matchShapes()	Contours
cv.createCLAHE()	Histograms
cv.dft()	Image Transforms
cv.getOptimalDFTSize()	Image Transforms
cv.HoughLines()	Hough Transform
cv.HoughLinesP()	Hough Transform
cv.HoughCircles()	Hough Transform

[huningxin]

@ganwenyao , thanks for reporting this issue. Could you please specify the usages or references to python tutorials which are using these functions?

[ganwenyao]

@huningxin

• cv.getRotationMatrix2D() can find an affine matrix of 2D rotation.
• cv.getPerspectiveTransform() can find the transformation matrix from 4 points on the input image and corresponding points on the output image.
• cv.filter2D() can convolve an arbitrary kernel with an image.
• cv.morphologyEx() can do Opening, Closing, Morphological Gradient and Top Hat morphological operation.
• cv.getStructuringElement() can generate elliptical/circular shaped kernels in morphological operation.
• cv.convertScaleAbs() can calculates absolute values, and converts the result to 8-bit.
• cv.isContourConvex() can check if a curve is convex or not.
• cv.minAreaRect() finds a rotated rectangle of the minimum area.
• cv.minEnclosingCircle() can find the circumcircle of an object.
• cv.fitLine() can fit a line to a set of points.
• cv.minMaxLoc() can finds the global minimum and maximum in an array.
• cv.convexityDefects() can finds the convexity defects of a contour.
• cv.pointPolygonTest() can finds the shortest distance between a point in the image and a contour.
• cv.matchShapes() enables us to compare two shapes, or two contours and returns a metric showing the similarity.
• cv.createCLAHE() is Contrast Limited Adaptive Histogram Equalization.
• cv.dft(): Fourier Transform
• cv.getOptimalDFTSize() returns the optimal DFT size for a given vector size.
• cv.HoughLines() cv.HoughLinesP() cv.HoughCircles(): Hough Transform

[huningxin]

As discussed in sync meeting, we will expose them. Let's follow up like this:

1. @huningxin will implement them in binding and report the file size difference.
2. @ganwenyao , please mark these functions in the google spreadsheet with specific color and notes in a column.
3. @sajjadt , please follow up with Vadim about these new functions.

[huningxin]

opencv.js size without and with these functions: 7039927 vs. 7211107 bytes

It increases about 167 KB.

[huningxin]

minEnclosingCircle is tricky, as http://docs.opencv.org/3.2.0/d3/dc0/group__imgproc__shape.html#ga8ce13c24081bbc7151e9326f412190f1, the third parameter is float& which returns the radius. However, according to https://github.com/kripken/emscripten/issues/611, embind doesn't support the non-const ref out parameters.

We may need to provide a wrapper to change the API signature in JS. Track it by https://github.com/huningxin/opencv/issues/126

[huningxin]

Similar issue to minMaxLoc http://docs.opencv.org/3.2.0/d2/de8/group__core__array.html#gab473bf2eb6d14ff97e89b355dac20707, it uses double* as return value. It's unfriendly for JS developer to provide the emscripten pointer to this API.

We may need to provide a wrapper to change the API signature in JS. Track it by https://github.com/huningxin/opencv/issues/127

[huningxin]

createCLAHE will be exposed as constructor of CLAHE object.

[ganwenyao]

These functions are marked with red color in the following google spreadsheet: https://docs.google.com/spreadsheets/d/1BGsasC0Afd_sS2lljt7GVx1cl8VUwAfBlVtddtXsfYM/edit?usp=sharing