英文:
Convolution in Golang
问题
我想在一张图像上执行卷积运算。
原始图像如下：
所以我用GIMP测试了卷积运算。使用以下矩阵：
1 1 1
1 1 1
1 1 1
和除数9
我得到了以下结果：
当我执行我的算法时，得到了以下结果：
我的算法是：
func Convolution(img *image.Image, matrice [][]int) *image.NRGBA {
imageRGBA := image.NewNRGBA((*img).Bounds())
w := (*img).Bounds().Dx()
h := (*img).Bounds().Dy()
sumR := 0
sumB := 0
sumG := 0
var r uint32
var g uint32
var b uint32
for y := 0; y < h; y++ {
for x := 0; x < w; x++ {
for i := -1; i <= 1; i++ {
for j := -1; j <= 1; j++ {
var imageX int
var imageY int
imageX = x + i
imageY = y + j
r, g, b, _ = (*img).At(imageX, imageY).RGBA()
sumG = (sumG + (int(g) * matrice[i+1][j+1]))
sumR = (sumR + (int(r) * matrice[i+1][j+1]))
sumB = (sumB + (int(b) * matrice[i+1][j+1]))
}
}
imageRGBA.Set(x, y, color.NRGBA{
uint8(min(sumR/9, 255)),
uint8(min(sumG/9, 255)),
uint8(min(sumB/9, 255)),
255,
})
sumR = 0
sumB = 0
sumG = 0
}
}
return imageRGBA
}
哪里出错了？
谢谢你的帮助。
英文:
I want execute a convolution product on an image.
The original image is:
So I test the convolution with gimp. With this matrix:
1 1 1
1 1 1
1 1 1
and the divider 9
I obtain
When I execute my algorithm I obtain:
My algorithm is:
func Convolution(img *image.Image, matrice [][]int) *image.NRGBA {
imageRGBA := image.NewNRGBA((*img).Bounds())
w := (*img).Bounds().Dx()
h := (*img).Bounds().Dy()
sumR := 0
sumB := 0
sumG := 0
var r uint32
var g uint32
var b uint32
for y := 0; y < h; y++ {
for x := 0; x < w; x++ {
for i := -1; i <= 1; i++ {
for j := -1; j <= 1; j++ {
var imageX int
var imageY int
imageX = x + i
imageY = y + j
r, g, b, _ = (*img).At(imageX, imageY).RGBA()
sumG = (sumG + (int(g) * matrice[i+1][j+1]))
sumR = (sumR + (int(r) * matrice[i+1][j+1]))
sumB = (sumB + (int(b) * matrice[i+1][j+1]))
}
}
imageRGBA.Set(x, y, color.NRGBA{
uint8(min(sumR/9, 255)),
uint8(min(sumG/9, 255)),
uint8(min(sumB/9, 255)),
255,
})
sumR = 0
sumB = 0
sumG = 0
}
}
return imageRGBA
}
Where are the error ?
Thank you for your help.
答案1
得分: 2
r、g和b是uint32类型的值，它们包含16位的颜色信息，如果以非零的8位值开始，它们的值总是大于255。
因此，你不能对RGBA值进行操作并将它们截断为uint8类型，因为这样会得到一个无用的结果，因为最低有效位只是8位值的小数部分。
将候选整数值与最大的16位值65535进行比较，并在截断之前将其向左移动8位，以获取8个最高有效位。
uint8(min(sumR/9, 0xffff) >> 8),
uint8(min(sumG/9, 0xffff) >> 8),
uint8(min(sumB/9, 0xffff) >> 8),
英文:
r, g, and b are uint32 values, and they contain 16bits of color information which is always greater than 255 if started as a non-zero 8 bit value.
You then can't operate on the RGBA values and truncate them to a uint8; that gives you a useless result because the least significant bits are just fractional parts of the 8bit values.
Compare the candidate integer value with the max 16bit value 65535, and shift it 8 bits before truncating it to get the 8 most significant bits.
uint8(min(sumR/9, 0xffff) >> 8),
uint8(min(sumG/9, 0xffff) >> 8),
uint8(min(sumB/9, 0xffff) >> 8),
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论