// RunCluster runs the kmeans algorithm on the image data with K = clusters
// returning the best kmeans.AppIter, the new image and time take to run the algorithm
func RunCluster(img image.Image, clusters, runs int, parallel bool) (imgObj image.Image, iter *kmeans.AppIter, duration time.Duration) {
// Convert the image to rgba if not already
RGBAImage, ok := img.(*image.RGBA)
// if not RGBA, do the convertion
if !ok {
log.Println("Explicit convertion to RGBA")
RGBAImage = kimage.ConvertToRGBA(img)
}
// convert the RGBAImage to a matrix
imageMatrix := kimage.RGBAToMatrix(RGBAImage)
// create the kmeans app
app := kmeans.NewApp(imageMatrix, runs, clusters, 0)
startTime := time.Now()
if parallel {
app.ParallelRun()
} else {
app.Run()
}
duration = time.Now().Sub(startTime) // track time taken
// convert the matrix back to an image
iter = app.BestIter
matrix := iter.ReconstructXMatrix()
rawMatrixer := matrix.(mat64.RawMatrixer)
imgObj = kimage.MatrixToRGBA(rawMatrixer, RGBAImage.Bounds().Size())
return
}
// ClusterImage runs the kmeans algorithm on the image data with K = colors
// returning the best kmeans.AppIter
func ClusterImage(imgData mat64.Matrix, colors, runs int) *kmeans.AppIter {
app := kmeans.NewApp(imgData, runs, colors, 0)
app.ParallelRun()
return app.BestIter
}