第五次图像信息处理作业，大致上就是图像的模糊和锐化吧。
Assignment-5作业要求
- Image mean filtering
- Laplacian image enhancement
作业分析
均值滤波
其实由上次作业可以知道，图像的处理就是由一系列矩阵变换得到的结果，不过在这次作业中，我们选择的是对像素进行矩阵变换而不是像素坐标，而均值滤波，顾名思义，就是对图像像素取均值得到新的像素值，以$3\times3$矩阵为例，其变换矩阵如下
值得注意的是，对于边缘的像素值，我们在这里对其进行保留处理。由此我们可以很快的写出代码
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
| uint8_t* imgMean(const uint8_t* imgData, int bitCount, int height, int width) {
int lineBytes = (bitCount * width / 8 + 3) / 4 * 4;
uint8_t* meanData = new uint8_t[lineBytes * height]{};
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
if (i == 0 || i == height - 1 || j == 0 || j == width - 1) {
*(meanData + i * lineBytes + j * 3) = *(imgData + i * lineBytes + j * 3);
*(meanData + i * lineBytes + j * 3 + 1) = *(imgData + i * lineBytes + j * 3 + 1);
*(meanData + i * lineBytes + j * 3 + 2) = *(imgData + i * lineBytes + j * 3 + 2);
}
else {
int r = *(imgData + i * lineBytes + j * 3) + *(imgData + (i + 1) * lineBytes + j * 3) + *(imgData + (i - 1) * lineBytes + j * 3) +
*(imgData + i * lineBytes + (j - 1) * 3) + *(imgData + i * lineBytes + (j + 1) * 3) + *(imgData + (i - 1) * lineBytes + (j - 1) * 3) +
*(imgData + (i - 1) * lineBytes + (j + 1) * 3) + *(imgData + (i + 1) * lineBytes + (j - 1) * 3) + *(imgData + (i + 1) * lineBytes + (j + 1) * 3);
int g = *(imgData + i * lineBytes + j * 3 + 1) + *(imgData + (i + 1) * lineBytes + j * 3 + 1) + *(imgData + (i - 1) * lineBytes + j * 3 + 1) +
*(imgData + i * lineBytes + (j - 1) * 3 + 1) + *(imgData + i * lineBytes + (j + 1) * 3 + 1) + *(imgData + (i - 1) * lineBytes + (j - 1) * 3 + 1) +
*(imgData + (i - 1) * lineBytes + (j + 1) * 3 + 1) + *(imgData + (i + 1) * lineBytes + (j - 1) * 3 + 1) + *(imgData + (i + 1) * lineBytes + (j + 1) * 3 + 1);
int b = *(imgData + i * lineBytes + j * 3 + 2) + *(imgData + (i + 1) * lineBytes + j * 3 + 2) + *(imgData + (i - 1) * lineBytes + j * 3 + 2) +
*(imgData + i * lineBytes + (j - 1) * 3 + 2) + *(imgData + i * lineBytes + (j + 1) * 3 + 2) + *(imgData + (i - 1) * lineBytes + (j - 1) * 3 + 2) +
*(imgData + (i - 1) * lineBytes + (j + 1) * 3 + 2) + *(imgData + (i + 1) * lineBytes + (j - 1) * 3 + 2) + *(imgData + (i + 1) * lineBytes + (j + 1) * 3 + 2);
*(meanData + i * lineBytes + j * 3) = (uint8_t)(r / 9);
*(meanData + i * lineBytes + j * 3 + 1) = (uint8_t)(g / 9);
*(meanData + i * lineBytes + j * 3 + 2) = (uint8_t)(b / 9);
}
}
}
return meanData;
}
|
最终结果对比
当然我们也可以修改均值滤波的矩阵，例如改成如下
那么这样图像的亮度也会有些许变化
拉普拉斯图像增强
这里我们使用拉普拉斯算子进行图像的增强，其中我们需要先计算出拉普拉斯遮罩
然后我们将原像素减去拉普拉斯遮罩即可得出锐化之后的图像，当然这里需要注意增强后的图像溢色的问题
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
| uint8_t* imgLaplacian(const uint8_t* imgData, int bitCount, int height, int width) {
int lineBytes = (bitCount * width / 8 + 3) / 4 * 4;
uint8_t* lapData = new uint8_t[lineBytes * height]{};
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
if (i == 0 || i == height - 1 || j == 0 || j == width - 1) {
*(lapData + i * lineBytes + j * 3) = *(imgData + i * lineBytes + j * 3);
*(lapData + i * lineBytes + j * 3 + 1) = *(imgData + i * lineBytes + j * 3 + 1);
*(lapData + i * lineBytes + j * 3 + 2) = *(imgData + i * lineBytes + j * 3 + 2);
}
else {
int r = 8 * *(imgData + i * lineBytes + j * 3) - *(imgData + (i + 1) * lineBytes + j * 3) - *(imgData + (i - 1) * lineBytes + j * 3) -
*(imgData + i * lineBytes + (j - 1) * 3) - *(imgData + i * lineBytes + (j + 1) * 3) - *(imgData + (i - 1) * lineBytes + (j - 1) * 3) -
*(imgData + (i - 1) * lineBytes + (j + 1) * 3) - *(imgData + (i + 1) * lineBytes + (j - 1) * 3) - *(imgData + (i + 1) * lineBytes + (j + 1) * 3);
int g = 8 * *(imgData + i * lineBytes + j * 3 + 1) - *(imgData + (i + 1) * lineBytes + j * 3 + 1) - *(imgData + (i - 1) * lineBytes + j * 3 + 1) -
*(imgData + i * lineBytes + (j - 1) * 3 + 1) - *(imgData + i * lineBytes + (j + 1) * 3 + 1) - *(imgData + (i - 1) * lineBytes + (j - 1) * 3 + 1) -
*(imgData + (i - 1) * lineBytes + (j + 1) * 3 + 1) - *(imgData + (i + 1) * lineBytes + (j - 1) * 3 + 1) - *(imgData + (i + 1) * lineBytes + (j + 1) * 3 + 1);
int b = 8 * *(imgData + i * lineBytes + j * 3 + 2) - *(imgData + (i + 1) * lineBytes + j * 3 + 2) - *(imgData + (i - 1) * lineBytes + j * 3 + 2) -
*(imgData + i * lineBytes + (j - 1) * 3 + 2) - *(imgData + i * lineBytes + (j + 1) * 3 + 2) - *(imgData + (i - 1) * lineBytes + (j - 1) * 3 + 2) -
*(imgData + (i - 1) * lineBytes + (j + 1) * 3 + 2) - *(imgData + (i + 1) * lineBytes + (j - 1) * 3 + 2) - *(imgData + (i + 1) * lineBytes + (j + 1) * 3 + 2);
r += *(imgData + i * lineBytes + j * 3);
g += *(imgData + i * lineBytes + j * 3 + 1);
b += *(imgData + i * lineBytes + j * 3 + 2);
r = r > 255 ? 255 : r;
g = g > 255 ? 255 : g;
b = b > 255 ? 255 : b;
r = r < 0 ? 0 : r;
g = g < 0 ? 0 : g;
b = b < 0 ? 0 : b;
*(lapData + i * lineBytes + j * 3) = r;
*(lapData + i * lineBytes + j * 3 + 1) = g;
*(lapData + i * lineBytes + j * 3 + 2) = b;
}
}
}
return lapData;
}
|
最终结果如下
总结
这次作业确实还蛮简单的，最后总计的编程时间可能一个小时都不到，不过这些矩阵运算还是十分重要的
同时长远地想，配合别的不同的滤波器可以实现其他不同的效果，可以算是图像滤波的入门体验作业吧