这个素描滤镜的算法我是在网上找到的，具体的链接和作者信息忘记了。（侵权则删）
/**
* 素描效果
* @param bmp
* @return
*/
public static Bitmap convertToSketch(Bitmap bmp) {
int pos, row, col, clr;
int width = bmp.getWidth();
int height = bmp.getHeight();
int[] pixSrc = new int[width * height];
int[] pixNvt = new int[width * height];
// 先对图象的像素处理成灰度颜色后再取反
bmp.getPixels(pixSrc, 0, width, 0, 0, width, height);
for (row = 0; row < height; row++) {
for (col = 0; col < width; col++) {
pos = row * width + col;
pixSrc[pos] = (Color.red(pixSrc[pos])
+ Color.green(pixSrc[pos]) + Color.blue(pixSrc[pos])) / 3;
pixNvt[pos] = 255 - pixSrc[pos];
}
}
// 对取反的像素进行高斯模糊, 强度可以设置，暂定为5.0
gaussGray(pixNvt, 5.0, 5.0, width, height);
// 灰度颜色和模糊后像素进行差值运算
for (row = 0; row < height; row++) {
for (col = 0; col < width; col++) {
pos = row * width + col;
clr = pixSrc[pos] << 8;
clr /= 256 - pixNvt[pos];
clr = Math.min(clr, 255);
pixSrc[pos] = Color.rgb(clr, clr, clr);
}
}
bmp.setPixels(pixSrc, 0, width, 0, 0, width, height);
return bmp;
}
private static int gaussGray(int[] psrc, double horz, double vert,
int width, int height) {
int[] dst, src;
double[] n_p, n_m, d_p, d_m, bd_p, bd_m;
double[] val_p, val_m;
int i, j, t, k, row, col, terms;
int[] initial_p, initial_m;
double std_dev;
int row_stride = width;
int max_len = Math.max(width, height);
int sp_p_idx, sp_m_idx, vp_idx, vm_idx;
val_p = new double[max_len];
val_m = new double[max_len];
n_p = new double[5];
n_m = new double[5];
d_p = new double[5];
d_m = new double[5];
bd_p = new double[5];
bd_m = new double[5];
src = new int[max_len];
dst = new int[max_len];
initial_p = new int[4];
initial_m = new int[4];
// 垂直方向
if (vert > 0.0) {
vert = Math.abs(vert) + 1.0;
std_dev = Math.sqrt(-(vert * vert) / (2 * Math.log(1.0 / 255.0)));
// 初试化常量
findConstants(n_p, n_m, d_p, d_m, bd_p, bd_m, std_dev);
for (col = 0; col < width; col++) {
for (k = 0; k < max_len; k++) {
val_m[k] = val_p[k] = 0;
}
for (t = 0; t < height; t++) {
src[t] = psrc[t * row_stride + col];
}
sp_p_idx = 0;
sp_m_idx = height - 1;
vp_idx = 0;
vm_idx = height - 1;
initial_p[0] = src[0];
initial_m[0] = src[height - 1];
for (row = 0; row < height; row++) {
terms = (row < 4) ? row : 4;
for (i = 0; i <= terms; i++) {
val_p[vp_idx] += n_p[i] * src[sp_p_idx - i] - d_p[i]
* val_p[vp_idx - i];
val_m[vm_idx] += n_m[i] * src[sp_m_idx + i] - d_m[i]
* val_m[vm_idx + i];
}
for (j = i; j <= 4; j++) {
val_p[vp_idx] += (n_p[j] - bd_p[j]) * initial_p[0];
val_m[vm_idx] += (n_m[j] - bd_m[j]) * initial_m[0];
}
sp_p_idx++;
sp_m_idx--;
vp_idx++;
vm_idx--;
}
transferGaussPixels(val_p, val_m, dst, 1, height);
for (t = 0; t < height; t++) {
psrc[t * row_stride + col] = dst[t];
}
}
}
// 水平方向
if (horz > 0.0) {
horz = Math.abs(horz) + 1.0;
if (horz != vert) {
std_dev = Math.sqrt(-(horz * horz)
/ (2 * Math.log(1.0 / 255.0)));
// 初试化常量
findConstants(n_p, n_m, d_p, d_m, bd_p, bd_m, std_dev);
}
for (row = 0; row < height; row++) {
for (k = 0; k < max_len; k++) {
val_m[k] = val_p[k] = 0;
}
for (t = 0; t < width; t++) {
src[t] = psrc[row * row_stride + t];
}
sp_p_idx = 0;
sp_m_idx = width - 1;
vp_idx = 0;
vm_idx = width - 1;
initial_p[0] = src[0];
initial_m[0] = src[width - 1];
for (col = 0; col < width; col++) {
terms = (col < 4) ? col : 4;
for (i = 0; i <= terms; i++) {
val_p[vp_idx] += n_p[i] * src[sp_p_idx - i] - d_p[i]
* val_p[vp_idx - i];
val_m[vm_idx] += n_m[i] * src[sp_m_idx + i] - d_m[i]
* val_m[vm_idx + i];
}
for (j = i; j <= 4; j++) {
val_p[vp_idx] += (n_p[j] - bd_p[j]) * initial_p[0];
val_m[vm_idx] += (n_m[j] - bd_m[j]) * initial_m[0];
}
sp_p_idx++;
sp_m_idx--;
vp_idx++;
vm_idx--;
}
transferGaussPixels(val_p, val_m, dst, 1, width);
for (t = 0; t < width; t++) {
psrc[row * row_stride + t] = dst[t];
}
}
}
return 0;
}
private static void transferGaussPixels(double[] src1, double[] src2,
int[] dest, int bytes, int width) {
int i, j, k, b;
int bend = bytes * width;
double sum;
i = j = k = 0;
for (b = 0; b < bend; b++) {
sum = src1[i++] + src2[j++];
if (sum > 255)
sum = 255;
else if (sum < 0)
sum = 0;
dest[k++] = (int) sum;
}
}
private static void findConstants(double[] n_p, double[] n_m, double[] d_p,
double[] d_m, double[] bd_p, double[] bd_m, double std_dev) {
double div = Math.sqrt(2 * 3.141593) * std_dev;
double x0 = -1.783 / std_dev;
double x1 = -1.723 / std_dev;
double x2 = 0.6318 / std_dev;
double x3 = 1.997 / std_dev;
double x4 = 1.6803 / div;
double x5 = 3.735 / div;
double x6 = -0.6803 / div;
double x7 = -0.2598 / div;
int i;
n_p[0] = x4 + x6;
n_p[1] = (Math.exp(x1)
* (x7 * Math.sin(x3) - (x6 + 2 * x4) * Math.cos(x3)) + Math
.exp(x0) * (x5 * Math.sin(x2) - (2 * x6 + x4) * Math.cos(x2)));
n_p[2] = (2
* Math.exp(x0 + x1)
* ((x4 + x6) * Math.cos(x3) * Math.cos(x2) - x5 * Math.cos(x3)
* Math.sin(x2) - x7 * Math.cos(x2) * Math.sin(x3)) + x6
* Math.exp(2 * x0) + x4 * Math.exp(2 * x1));
n_p[3] = (Math.exp(x1 + 2 * x0)
* (x7 * Math.sin(x3) - x6 * Math.cos(x3)) + Math.exp(x0 + 2
* x1)
* (x5 * Math.sin(x2) - x4 * Math.cos(x2)));
n_p[4] = 0.0;
d_p[0] = 0.0;
d_p[1] = -2 * Math.exp(x1) * Math.cos(x3) - 2 * Math.exp(x0)
* Math.cos(x2);
d_p[2] = 4 * Math.cos(x3) * Math.cos(x2) * Math.exp(x0 + x1)
+ Math.exp(2 * x1) + Math.exp(2 * x0);
d_p[3] = -2 * Math.cos(x2) * Math.exp(x0 + 2 * x1) - 2 * Math.cos(x3)
* Math.exp(x1 + 2 * x0);
d_p[4] = Math.exp(2 * x0 + 2 * x1);
for (i = 0; i <= 4; i++) {
d_m[i] = d_p[i];
}
n_m[0] = 0.0;
for (i = 1; i <= 4; i++) {
n_m[i] = n_p[i] - d_p[i] * n_p[0];
}
double sum_n_p, sum_n_m, sum_d;
double a, b;
sum_n_p = 0.0;
sum_n_m = 0.0;
sum_d = 0.0;
for (i = 0; i <= 4; i++) {
sum_n_p += n_p[i];
sum_n_m += n_m[i];
sum_d += d_p[i];
}
a = sum_n_p / (1.0 + sum_d);
b = sum_n_m / (1.0 + sum_d);
for (i = 0; i <= 4; i++) {
bd_p[i] = d_p[i] * a;
bd_m[i] = d_m[i] * b;
}
}
其效果如下：
效果图 原图
作者：qq_32353771 发表于2017/1/5 19:45:25 原文链接
阅读：51 评论：0 查看评论