Initial commit

sobel.go

@@ -0,0 +1,105 @@
+package caire
+
+import (
+	"image"
+	"math"
+)
+
+type kernel [][]int32
+
+var (
+	kernelX kernel = kernel{
+		{-1, 0, 1},
+		{-2, 0, 2},
+		{-1, 0, 1},
+	}
+
+	kernelY kernel = kernel{
+		{-1, -2, -1},
+		{0, 0, 0},
+		{1, 2, 1},
+	}
+)
+
+// Detect image edges.
+// See https://en.wikipedia.org/wiki/Sobel_operator
+func SobelFilter(img *image.NRGBA, threshold float64) *image.NRGBA {
+	var sumX, sumY int32
+	dx, dy := img.Bounds().Max.X, img.Bounds().Max.Y
+	dst := image.NewNRGBA(img.Bounds())
+
+	// Get 3x3 window of pixels because image data given is just a 1D array of pixels
+	maxPixelOffset := dx*dy + len(kernelX) - 1
+
+	data := getImageData(img)
+	length := len(data) - maxPixelOffset
+	magnitudes := make([]int32, length)
+
+	for i := 0; i < length; i++ {
+		// Sum each pixel with the kernel value
+		sumX, sumY = 0, 0
+		for x := 0; x < len(kernelX); x++ {
+			for y := 0; y < len(kernelY); y++ {
+				px := data[i+(dx*y)+x]
+				if len(px) > 0 {
+					r := px[0]
+					// We are using px[0] (i.e. R value) because the image is grayscale anyway
+					sumX += int32(r) * kernelX[y][x]
+					sumY += int32(r) * kernelY[y][x]
+				}
+			}
+		}
+		magnitude := math.Sqrt(float64(sumX*sumX) + float64(sumY*sumY))
+		// Check for pixel color boundaries
+		if magnitude < 0 {
+			magnitude = 0
+		} else if magnitude > 255 {
+			magnitude = 255
+		}
+
+		// Set magnitude to 0 if doesn't exceed threshold, else set to magnitude
+		if magnitude > threshold {
+			magnitudes[i] = int32(magnitude)
+		} else {
+			magnitudes[i] = 0
+		}
+	}
+
+	dataLength := dx * dy * 4
+	edges := make([]int32, dataLength)
+
+	// Apply the kernel values.
+	for i := 0; i < dataLength; i++ {
+		if i%4 != 0 {
+			m := magnitudes[i/4]
+			if m != 0 {
+				edges[i-1] = m
+			}
+		}
+	}
+
+	// Generate the new image with the sobel filter applied.
+	for idx := 0; idx < len(edges); idx += 4 {
+		dst.Pix[idx] = uint8(edges[idx])
+		dst.Pix[idx+1] = uint8(edges[idx+1])
+		dst.Pix[idx+2] = uint8(edges[idx+2])
+		dst.Pix[idx+3] = 255
+	}
+	return dst
+}
+
+// Group pixels into 2D array, each one containing the pixel RGB value.
+func getImageData(img *image.NRGBA) [][]uint8 {
+	dx, dy := img.Bounds().Max.X, img.Bounds().Max.Y
+	pixels := make([][]uint8, dx*dy*4)
+
+	for i := 0; i < len(pixels); i += 4 {
+		pixels[i/4] = []uint8{
+			img.Pix[i],
+			img.Pix[i+1],
+			img.Pix[i+2],
+			img.Pix[i+3],
+		}
+	}
+	return pixels
+}