圖片JNI（C++\Java）高斯模糊 多線程
在我的博客中，曾經發布了一篇高斯模糊（堆棧模糊）的文章；在其中使用了國外的一個堆棧模糊來實現對圖片的模糊處理；同時弄了一個JNI C++ 的版本。
這篇文章依然是堆棧模糊；可以說最原始的地方還是堆棧模糊部分；只不過是支持多線程的。
納尼？？感情是之前那個不支持多線程？Sorry，我說錯了；兩個都是支持多線程調用的。不過新講的這個是能在內部采用多線程進行分段模糊。
原來的：[Android]-圖片JNI（C++\Java）高斯模糊的實現與比較
開工吧
說明：其中代碼大部分來源于網絡，不過都是開源的。
最原始的代碼：
stackblur.cpp
// The Stack Blur Algorithm was invented by Mario Klingemann,
// mario@quasimondo.com and described here:
// http://incubator.quasimondo.com/processing/fast_blur_deluxe.php
// This is C++ RGBA (32 bit color) multi-threaded version
// by Victor Laskin (victor.laskin@gmail.com)
// More details: http://vitiy.info/stackblur-algorithm-multi-threaded-blur-for-cpp
// This code is using MVThread class from my cross-platform framework
// You can exchange it with any thread implementation you like
// -------------------------------------- stackblur ----------------------------------------->
static unsigned short const stackblur_mul[255] =
{
512,512,456,512,328,456,335,512,405,328,271,456,388,335,292,512,
454,405,364,328,298,271,496,456,420,388,360,335,312,292,273,512,
482,454,428,405,383,364,345,328,312,298,284,271,259,496,475,456,
437,420,404,388,374,360,347,335,323,312,302,292,282,273,265,512,
497,482,468,454,441,428,417,405,394,383,373,364,354,345,337,328,
320,312,305,298,291,284,278,271,265,259,507,496,485,475,465,456,
446,437,428,420,412,404,396,388,381,374,367,360,354,347,341,335,
329,323,318,312,307,302,297,292,287,282,278,273,269,265,261,512,
505,497,489,482,475,468,461,454,447,441,435,428,422,417,411,405,
399,394,389,383,378,373,368,364,359,354,350,345,341,337,332,328,
324,320,316,312,309,305,301,298,294,291,287,284,281,278,274,271,
268,265,262,259,257,507,501,496,491,485,480,475,470,465,460,456,
451,446,442,437,433,428,424,420,416,412,408,404,400,396,392,388,
385,381,377,374,370,367,363,360,357,354,350,347,344,341,338,335,
332,329,326,323,320,318,315,312,310,307,304,302,299,297,294,292,
289,287,285,282,280,278,275,273,271,269,267,265,263,261,259
};
static unsigned char const stackblur_shr[255] =
{
9, 11, 12, 13, 13, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16, 17,
17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 18, 19,
19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 20, 20, 20,
20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 21,
21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21,
21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22,
22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22,
22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23,
23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23,
23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23,
23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23,
23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24
};
/// Stackblur algorithm body
void stackblurJob(unsigned char* src, ///< input image data
unsigned int w, ///< image width
unsigned int h, ///< image height
unsigned int radius, ///< blur intensity (should be in 2..254 range)
int cores, ///< total number of working threads
int core, ///< current thread number
int step, ///< step of processing (1,2)
unsigned char* stack ///< stack buffer
)
{
unsigned int x, y, xp, yp, i;
unsigned int sp;
unsigned int stack_start;
unsigned char* stack_ptr;
unsigned char* src_ptr;
unsigned char* dst_ptr;
unsigned long sum_r;
unsigned long sum_g;
unsigned long sum_b;
unsigned long sum_a;
unsigned long sum_in_r;
unsigned long sum_in_g;
unsigned long sum_in_b;
unsigned long sum_in_a;
unsigned long sum_out_r;
unsigned long sum_out_g;
unsigned long sum_out_b;
unsigned long sum_out_a;
unsigned int wm = w - 1;
unsigned int hm = h - 1;
unsigned int w4 = w * 4;
unsigned int div = (radius * 2) + 1;
unsigned int mul_sum = stackblur_mul[radius];
unsigned char shr_sum = stackblur_shr[radius];
if (step == 1)
{
int minY = core * h / cores;
int maxY = (core + 1) * h / cores;
for(y = minY; y < maxY; y++)
{
sum_r = sum_g = sum_b = sum_a =
sum_in_r = sum_in_g = sum_in_b = sum_in_a =
sum_out_r = sum_out_g = sum_out_b = sum_out_a = 0;
src_ptr = src + w4 * y; // start of line (0,y)
for(i = 0; i <= radius; i++)
{
stack_ptr = &stack[ 4 * i ];
stack_ptr[0] = src_ptr[0];
stack_ptr[1] = src_ptr[1];
stack_ptr[2] = src_ptr[2];
stack_ptr[3] = src_ptr[3];
sum_r += src_ptr[0] * (i + 1);
sum_g += src_ptr[1] * (i + 1);
sum_b += src_ptr[2] * (i + 1);
sum_a += src_ptr[3] * (i + 1);
sum_out_r += src_ptr[0];
sum_out_g += src_ptr[1];
sum_out_b += src_ptr[2];
sum_out_a += src_ptr[3];
}
for(i = 1; i <= radius; i++)
{
if (i <= wm) src_ptr += 4;
stack_ptr = &stack[ 4 * (i + radius) ];
stack_ptr[0] = src_ptr[0];
stack_ptr[1] = src_ptr[1];
stack_ptr[2] = src_ptr[2];
stack_ptr[3] = src_ptr[3];
sum_r += src_ptr[0] * (radius + 1 - i);
sum_g += src_ptr[1] * (radius + 1 - i);
sum_b += src_ptr[2] * (radius + 1 - i);
sum_a += src_ptr[3] * (radius + 1 - i);
sum_in_r += src_ptr[0];
sum_in_g += src_ptr[1];
sum_in_b += src_ptr[2];
sum_in_a += src_ptr[3];
}
sp = radius;
xp = radius;
if (xp > wm) xp = wm;
src_ptr = src + 4 * (xp + y * w); // img.pix_ptr(xp, y);
dst_ptr = src + y * w4; // img.pix_ptr(0, y);
for(x = 0; x < w; x++)
{
dst_ptr[0] = (sum_r * mul_sum) >> shr_sum;
dst_ptr[1] = (sum_g * mul_sum) >> shr_sum;
dst_ptr[2] = (sum_b * mul_sum) >> shr_sum;
dst_ptr[3] = (sum_a * mul_sum) >> shr_sum;
dst_ptr += 4;
sum_r -= sum_out_r;
sum_g -= sum_out_g;
sum_b -= sum_out_b;
sum_a -= sum_out_a;
stack_start = sp + div - radius;
if (stack_start >= div) stack_start -= div;
stack_ptr = &stack[4 * stack_start];
sum_out_r -= stack_ptr[0];
sum_out_g -= stack_ptr[1];
sum_out_b -= stack_ptr[2];
sum_out_a -= stack_ptr[3];
if(xp < wm)
{
src_ptr += 4;
++xp;
}
stack_ptr[0] = src_ptr[0];
stack_ptr[1] = src_ptr[1];
stack_ptr[2] = src_ptr[2];
stack_ptr[3] = src_ptr[3];
sum_in_r += src_ptr[0];
sum_in_g += src_ptr[1];
sum_in_b += src_ptr[2];
sum_in_a += src_ptr[3];
sum_r += sum_in_r;
sum_g += sum_in_g;
sum_b += sum_in_b;
sum_a += sum_in_a;
++sp;
if (sp >= div) sp = 0;
stack_ptr = &stack[sp*4];
sum_out_r += stack_ptr[0];
sum_out_g += stack_ptr[1];
sum_out_b += stack_ptr[2];
sum_out_a += stack_ptr[3];
sum_in_r -= stack_ptr[0];
sum_in_g -= stack_ptr[1];
sum_in_b -= stack_ptr[2];
sum_in_a -= stack_ptr[3];
}
}
}
// step 2
if (step == 2)
{
int minX = core * w / cores;
int maxX = (core + 1) * w / cores;
for(x = minX; x < maxX; x++)
{
sum_r = sum_g = sum_b = sum_a =
sum_in_r = sum_in_g = sum_in_b = sum_in_a =
sum_out_r = sum_out_g = sum_out_b = sum_out_a = 0;
src_ptr = src + 4 * x; // x,0
for(i = 0; i <= radius; i++)
{
stack_ptr = &stack[i * 4];
stack_ptr[0] = src_ptr[0];
stack_ptr[1] = src_ptr[1];
stack_ptr[2] = src_ptr[2];
stack_ptr[3] = src_ptr[3];
sum_r += src_ptr[0] * (i + 1);
sum_g += src_ptr[1] * (i + 1);
sum_b += src_ptr[2] * (i + 1);
sum_a += src_ptr[3] * (i + 1);
sum_out_r += src_ptr[0];
sum_out_g += src_ptr[1];
sum_out_b += src_ptr[2];
sum_out_a += src_ptr[3];
}
for(i = 1; i <= radius; i++)
{
if(i <= hm) src_ptr += w4; // +stride
stack_ptr = &stack[4 * (i + radius)];
stack_ptr[0] = src_ptr[0];
stack_ptr[1] = src_ptr[1];
stack_ptr[2] = src_ptr[2];
stack_ptr[3] = src_ptr[3];
sum_r += src_ptr[0] * (radius + 1 - i);
sum_g += src_ptr[1] * (radius + 1 - i);
sum_b += src_ptr[2] * (radius + 1 - i);
sum_a += src_ptr[3] * (radius + 1 - i);
sum_in_r += src_ptr[0];
sum_in_g += src_ptr[1];
sum_in_b += src_ptr[2];
sum_in_a += src_ptr[3];
}
sp = radius;
yp = radius;
if (yp > hm) yp = hm;
src_ptr = src + 4 * (x + yp * w); // img.pix_ptr(x, yp);
dst_ptr = src + 4 * x; // img.pix_ptr(x, 0);
for(y = 0; y < h; y++)
{
dst_ptr[0] = (sum_r * mul_sum) >> shr_sum;
dst_ptr[1] = (sum_g * mul_sum) >> shr_sum;
dst_ptr[2] = (sum_b * mul_sum) >> shr_sum;
dst_ptr[3] = (sum_a * mul_sum) >> shr_sum;
dst_ptr += w4;
sum_r -= sum_out_r;
sum_g -= sum_out_g;
sum_b -= sum_out_b;
sum_a -= sum_out_a;
stack_start = sp + div - radius;
if(stack_start >= div) stack_start -= div;
stack_ptr = &stack[4 * stack_start];
sum_out_r -= stack_ptr[0];
sum_out_g -= stack_ptr[1];
sum_out_b -= stack_ptr[2];
sum_out_a -= stack_ptr[3];
if(yp < hm)
{
src_ptr += w4; // stride
++yp;
}
stack_ptr[0] = src_ptr[0];
stack_ptr[1] = src_ptr[1];
stack_ptr[2] = src_ptr[2];
stack_ptr[3] = src_ptr[3];
sum_in_r += src_ptr[0];
sum_in_g += src_ptr[1];
sum_in_b += src_ptr[2];
sum_in_a += src_ptr[3];
sum_r += sum_in_r;
sum_g += sum_in_g;
sum_b += sum_in_b;
sum_a += sum_in_a;
++sp;
if (sp >= div) sp = 0;
stack_ptr = &stack[sp*4];
sum_out_r += stack_ptr[0];
sum_out_g += stack_ptr[1];
sum_out_b += stack_ptr[2];
sum_out_a += stack_ptr[3];
sum_in_r -= stack_ptr[0];
sum_in_g -= stack_ptr[1];
sum_in_b -= stack_ptr[2];
sum_in_a -= stack_ptr[3];
}
}
}
}
class MVImageUtilsStackBlurTask : public MVThread
{
public:
unsigned char* src;
unsigned int w;
unsigned int h;
unsigned int radius;
int cores;
int core;
int step;
unsigned char* stack;
inline MVImageUtilsStackBlurTask(unsigned char* src, unsigned int w, unsigned int h, unsigned int radius, int cores, int core, int step, unsigned char* stack)
{
this->src = src;
this->w = w;
this->h = h;
this->radius = radius;
this->cores = cores;
this->core = core;
this->step = step;
this->stack = stack;
}
inline void run()
{
stackblurJob(src, w, h, radius, cores, core, step, stack);
}
};
/// Stackblur algorithm by Mario Klingemann
/// Details here:
/// http://www.quasimondo.com/StackBlurForCanvas/StackBlurDemo.html
/// C++ implemenation base from:
/// https://gist.github.com/benjamin9999/3809142
/// http://www.antigrain.com/__code/include/agg_blur.h.html
/// This version works only with RGBA color
void stackblur(unsigned char* src, ///< input image data
unsigned int w, ///< image width
unsigned int h, ///< image height
unsigned int radius, ///< blur intensity (should be in 2..254 range)
int cores = 1 ///< number of threads (1 - normal single thread)
)
{
if (radius > 254) return;
if (radius < 2) return;
unsigned int div = (radius * 2) + 1;
unsigned char* stack = new unsigned char [div * 4 * cores];
if (cores == 1)
{
// no multithreading
stackblurJob(src, w, h, radius, 1, 0, 1, stack);
stackblurJob(src, w, h, radius, 1, 0, 2, stack);
}
else
{
MVImageUtilsStackBlurTask** workers = new MVImageUtilsStackBlurTask*[cores];
for (int i = 0; i < cores; i++)
{
workers[i] = new MVImageUtilsStackBlurTask(src, w, h, radius, cores, i, 1, stack + div * 4 * i);
workers[i]->start();
}
for (int i = 0; i < cores; i++)
workers[i]->wait();
for (int i = 0; i < cores; i++)
{
workers[i]->step = 2;
workers[i]->start();
}
for (int i = 0; i < cores; i++)
{
workers[i]->wait();
delete workers[i];
}
delete[] workers;
}
delete[] stack;
} 這個是C++的代碼。
在網上發現了一個開源的 Android 模糊的例子，其中有該代碼的 Java 版本。
JavaBlurProcess.java
import android.graphics.Bitmap;
import java.util.ArrayList;
import java.util.concurrent.Callable;
class JavaBlurProcess implements BlurProcess {
private static final short[] stackblur_mul = {
512, 512, 456, 512, 328, 456, 335, 512, 405, 328, 271, 456, 388, 335, 292, 512,
454, 405, 364, 328, 298, 271, 496, 456, 420, 388, 360, 335, 312, 292, 273, 512,
482, 454, 428, 405, 383, 364, 345, 328, 312, 298, 284, 271, 259, 496, 475, 456,
437, 420, 404, 388, 374, 360, 347, 335, 323, 312, 302, 292, 282, 273, 265, 512,
497, 482, 468, 454, 441, 428, 417, 405, 394, 383, 373, 364, 354, 345, 337, 328,
320, 312, 305, 298, 291, 284, 278, 271, 265, 259, 507, 496, 485, 475, 465, 456,
446, 437, 428, 420, 412, 404, 396, 388, 381, 374, 367, 360, 354, 347, 341, 335,
329, 323, 318, 312, 307, 302, 297, 292, 287, 282, 278, 273, 269, 265, 261, 512,
505, 497, 489, 482, 475, 468, 461, 454, 447, 441, 435, 428, 422, 417, 411, 405,
399, 394, 389, 383, 378, 373, 368, 364, 359, 354, 350, 345, 341, 337, 332, 328,
324, 320, 316, 312, 309, 305, 301, 298, 294, 291, 287, 284, 281, 278, 274, 271,
268, 265, 262, 259, 257, 507, 501, 496, 491, 485, 480, 475, 470, 465, 460, 456,
451, 446, 442, 437, 433, 428, 424, 420, 416, 412, 408, 404, 400, 396, 392, 388,
385, 381, 377, 374, 370, 367, 363, 360, 357, 354, 350, 347, 344, 341, 338, 335,
332, 329, 326, 323, 320, 318, 315, 312, 310, 307, 304, 302, 299, 297, 294, 292,
289, 287, 285, 282, 280, 278, 275, 273, 271, 269, 267, 265, 263, 261, 259
};
private static final byte[] stackblur_shr = {
9, 11, 12, 13, 13, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16, 17,
17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 18, 19,
19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 19, 20, 20, 20,
20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 21,
21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21,
21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22,
22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22,
22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23,
23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23,
23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23,
23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23, 23,
23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24
};
@Override
public Bitmap blur(Bitmap original, float radius) {
int w = original.getWidth();
int h = original.getHeight();
int[] currentPixels = new int[w * h];
original.getPixels(currentPixels, 0, w, 0, 0, w, h);
int cores = StackBlurManager.EXECUTOR_THREADS;
ArrayList<BlurTask> horizontal = new ArrayList<BlurTask>(cores);
ArrayList<BlurTask> vertical = new ArrayList<BlurTask>(cores);
for (int i = 0; i < cores; i++) {
horizontal.add(new BlurTask(currentPixels, w, h, (int) radius, cores, i, 1));
vertical.add(new BlurTask(currentPixels, w, h, (int) radius, cores, i, 2));
}
try {
StackBlurManager.EXECUTOR.invokeAll(horizontal);
} catch (InterruptedException e) {
return null;
}
try {
StackBlurManager.EXECUTOR.invokeAll(vertical);
} catch (InterruptedException e) {
return null;
}
return Bitmap.createBitmap(currentPixels, w, h, Bitmap.Config.ARGB_8888);
}
private static void blurIteration(int[] src, int w, int h, int radius, int cores, int core, int step) {
int x, y, xp, yp, i;
int sp;
int stack_start;
int stack_i;
int src_i;
int dst_i;
long sum_r, sum_g, sum_b,
sum_in_r, sum_in_g, sum_in_b,
sum_out_r, sum_out_g, sum_out_b;
int wm = w - 1;
int hm = h - 1;
int div = (radius * 2) + 1;
int mul_sum = stackblur_mul[radius];
byte shr_sum = stackblur_shr[radius];
int[] stack = new int[div];
if (step == 1)
{
int minY = core * h / cores;
int maxY = (core + 1) * h / cores;
for(y = minY; y < maxY; y++)
{
sum_r = sum_g = sum_b =
sum_in_r = sum_in_g = sum_in_b =
sum_out_r = sum_out_g = sum_out_b = 0;
src_i = w * y; // start of line (0,y)
for(i = 0; i <= radius; i++)
{
stack_i = i;
stack[stack_i] = src[src_i];
sum_r += ((src[src_i] >>> 16) & 0xff) * (i + 1);
sum_g += ((src[src_i] >>> 8) & 0xff) * (i + 1);
sum_b += (src[src_i] & 0xff) * (i + 1);
sum_out_r += ((src[src_i] >>> 16) & 0xff);
sum_out_g += ((src[src_i] >>> 8) & 0xff);
sum_out_b += (src[src_i] & 0xff);
}
for(i = 1; i <= radius; i++)
{
if (i <= wm) src_i += 1;
stack_i = i + radius;
stack[stack_i] = src[src_i];
sum_r += ((src[src_i] >>> 16) & 0xff) * (radius + 1 - i);
sum_g += ((src[src_i] >>> 8) & 0xff) * (radius + 1 - i);
sum_b += (src[src_i] & 0xff) * (radius + 1 - i);
sum_in_r += ((src[src_i] >>> 16) & 0xff);
sum_in_g += ((src[src_i] >>> 8) & 0xff);
sum_in_b += (src[src_i] & 0xff);
}
sp = radius;
xp = radius;
if (xp > wm) xp = wm;
src_i = xp + y * w; // img.pix_ptr(xp, y);
dst_i = y * w; // img.pix_ptr(0, y);
for(x = 0; x < w; x++)
{
src[dst_i] = (int)
((src[dst_i] & 0xff000000) |
((((sum_r * mul_sum) >>> shr_sum) & 0xff) << 16) |
((((sum_g * mul_sum) >>> shr_sum) & 0xff) << 8) |
((((sum_b * mul_sum) >>> shr_sum) & 0xff)));
dst_i += 1;
sum_r -= sum_out_r;
sum_g -= sum_out_g;
sum_b -= sum_out_b;
stack_start = sp + div - radius;
if (stack_start >= div) stack_start -= div;
stack_i = stack_start;
sum_out_r -= ((stack[stack_i] >>> 16) & 0xff);
sum_out_g -= ((stack[stack_i] >>> 8) & 0xff);
sum_out_b -= (stack[stack_i] & 0xff);
if(xp < wm)
{
src_i += 1;
++xp;
}
stack[stack_i] = src[src_i];
sum_in_r += ((src[src_i] >>> 16) & 0xff);
sum_in_g += ((src[src_i] >>> 8) & 0xff);
sum_in_b += (src[src_i] & 0xff);
sum_r += sum_in_r;
sum_g += sum_in_g;
sum_b += sum_in_b;
++sp;
if (sp >= div) sp = 0;
stack_i = sp;
sum_out_r += ((stack[stack_i] >>> 16) & 0xff);
sum_out_g += ((stack[stack_i] >>> 8) & 0xff);
sum_out_b += (stack[stack_i] & 0xff);
sum_in_r -= ((stack[stack_i] >>> 16) & 0xff);
sum_in_g -= ((stack[stack_i] >>> 8) & 0xff);
sum_in_b -= (stack[stack_i] & 0xff);
}
}
}
// step 2
else if (step == 2)
{
int minX = core * w / cores;
int maxX = (core + 1) * w / cores;
for(x = minX; x < maxX; x++)
{
sum_r = sum_g = sum_b =
sum_in_r = sum_in_g = sum_in_b =
sum_out_r = sum_out_g = sum_out_b = 0;
src_i = x; // x,0
for(i = 0; i <= radius; i++)
{
stack_i = i;
stack[stack_i] = src[src_i];
sum_r += ((src[src_i] >>> 16) & 0xff) * (i + 1);
sum_g += ((src[src_i] >>> 8) & 0xff) * (i + 1);
sum_b += (src[src_i] & 0xff) * (i + 1);
sum_out_r += ((src[src_i] >>> 16) & 0xff);
sum_out_g += ((src[src_i] >>> 8) & 0xff);
sum_out_b += (src[src_i] & 0xff);
}
for(i = 1; i <= radius; i++)
{
if(i <= hm) src_i += w; // +stride
stack_i = i + radius;
stack[stack_i] = src[src_i];
sum_r += ((src[src_i] >>> 16) & 0xff) * (radius + 1 - i);
sum_g += ((src[src_i] >>> 8) & 0xff) * (radius + 1 - i);
sum_b += (src[src_i] & 0xff) * (radius + 1 - i);
sum_in_r += ((src[src_i] >>> 16) & 0xff);
sum_in_g += ((src[src_i] >>> 8) & 0xff);
sum_in_b += (src[src_i] & 0xff);
}
sp = radius;
yp = radius;
if (yp > hm) yp = hm;
src_i = x + yp * w; // img.pix_ptr(x, yp);
dst_i = x; // img.pix_ptr(x, 0);
for(y = 0; y < h; y++)
{
src[dst_i] = (int)
((src[dst_i] & 0xff000000) |
((((sum_r * mul_sum) >>> shr_sum) & 0xff) << 16) |
((((sum_g * mul_sum) >>> shr_sum) & 0xff) << 8) |
((((sum_b * mul_sum) >>> shr_sum) & 0xff)));
dst_i += w;
sum_r -= sum_out_r;
sum_g -= sum_out_g;
sum_b -= sum_out_b;
stack_start = sp + div - radius;
if(stack_start >= div) stack_start -= div;
stack_i = stack_start;
sum_out_r -= ((stack[stack_i] >>> 16) & 0xff);
sum_out_g -= ((stack[stack_i] >>> 8) & 0xff);
sum_out_b -= (stack[stack_i] & 0xff);
if(yp < hm)
{
src_i += w; // stride
++yp;
}
stack[stack_i] = src[src_i];
sum_in_r += ((src[src_i] >>> 16) & 0xff);
sum_in_g += ((src[src_i] >>> 8) & 0xff);
sum_in_b += (src[src_i] & 0xff);
sum_r += sum_in_r;
sum_g += sum_in_g;
sum_b += sum_in_b;
++sp;
if (sp >= div) sp = 0;
stack_i = sp;
sum_out_r += ((stack[stack_i] >>> 16) & 0xff);
sum_out_g += ((stack[stack_i] >>> 8) & 0xff);
sum_out_b += (stack[stack_i] & 0xff);
sum_in_r -= ((stack[stack_i] >>> 16) & 0xff);
sum_in_g -= ((stack[stack_i] >>> 8) & 0xff);
sum_in_b -= (stack[stack_i] & 0xff);
}
}
}
}
private static class BlurTask implements Callable<Void> {
private final int[] _src;
private final int _w;
private final int _h;
private final int _radius;
private final int _totalCores;
private final int _coreIndex;
private final int _round;
public BlurTask(int[] src, int w, int h, int radius, int totalCores, int coreIndex, int round) {
_src = src;
_w = w;
_h = h;
_radius = radius;
_totalCores = totalCores;
_coreIndex = coreIndex;
_round = round;
}
@Override public Void call() throws Exception {
blurIteration(_src, _w, _h, _radius, _totalCores, _coreIndex, _round);
return null;
}
}
} 該代碼與C++的代碼同出一源。下面來簡要分析一下。
分析
入口：
public Bitmap blur(Bitmap original, float radius) {
int w = original.getWidth();
int h = original.getHeight();
int[] currentPixels = new int[w * h];
original.getPixels(currentPixels, 0, w, 0, 0, w, h);
int cores = StackBlurManager.EXECUTOR_THREADS;
ArrayList<BlurTask> horizontal = new ArrayList<BlurTask>(cores);
ArrayList<BlurTask> vertical = new ArrayList<BlurTask>(cores);
for (int i = 0; i < cores; i++) {
horizontal.add(new BlurTask(currentPixels, w, h, (int) radius, cores, i, 1));
vertical.add(new BlurTask(currentPixels, w, h, (int) radius, cores, i, 2));
}
try {
StackBlurManager.EXECUTOR.invokeAll(horizontal);
} catch (InterruptedException e) {
return null;
}
try {
StackBlurManager.EXECUTOR.invokeAll(vertical);
} catch (InterruptedException e) {
return null;
}
return Bitmap.createBitmap(currentPixels, w, h, Bitmap.Config.ARGB_8888);
} 方法進入后，首先得到圖片的像素點集合；而后
New 兩個
List ，分別存儲
BlurTask 。
看看 BlurTask ：
private static class BlurTask implements Callable<Void> {
private final int[] _src;
private final int _w;
private final int _h;
private final int _radius;
private final int _totalCores;
private final int _coreIndex;
private final int _round;
public BlurTask(int[] src, int w, int h, int radius, int totalCores, int coreIndex, int round) {
_src = src;
_w = w;
_h = h;
_radius = radius;
_totalCores = totalCores;
_coreIndex = coreIndex;
_round = round;
}
@Override public Void call() throws Exception {
blurIteration(_src, _w, _h, _radius, _totalCores, _coreIndex, _round);
return null;
}
} 可以看出
BlurTask 是用于進行
線程池的回調類；其主要調用了方法
blurIteration進行模糊。
同時給方法，傳入的參數有：
- 像素點集合
- 寬
- 高
- 模糊半徑
- 最大內核數
- 當前內核
步驟
而后回到第一步處：
int cores = StackBlurManager.EXECUTOR_THREADS;其實是調用的：
Runtime.getRuntime().availableProcessors();該方法用于返回當前虛擬機的核數；用此方法可以 最大效率的進行多線程的分配。
而后：
for (int i = 0; i < cores; i++) {
horizontal.add(new BlurTask(currentPixels, w, h, (int) radius, cores, i, 1));
vertical.add(new BlurTask(currentPixels, w, h, (int) radius, cores, i, 2));
} 可以看出往列表中加上，對應的線程回調操作；分別是
步驟1，和步驟2
而后：
try {
StackBlurManager.EXECUTOR.invokeAll(horizontal);
} catch (InterruptedException e) {
return null;
}
try {
StackBlurManager.EXECUTOR.invokeAll(vertical);
} catch (InterruptedException e) {
return null;
} 該方法中：
StackBlurManager.EXECUTOR 其實是:
Executors.newFixedThreadPool(Runtime.getRuntime().availableProcessors());可以看出其實是建立了一個線程池，線程池的運行線程數為當前虛擬機的核數。
然后把步驟1的集合丟進線程池并等待執行完成；而后丟進步驟2.
其實從其命名就能看出：
- 步驟1其實是模糊水平方向
- 步驟2其實是模糊垂直方向
具體是不是可以看看 blurIteration 方法。
由于方法太長就簡化了一下：
private static void blurIteration(int[] src, int w, int h, int radius, int cores, int core, int step) {
...
if (step == 1)
{
int minY = core * h / cores;
int maxY = (core + 1) * h / cores;
for(y = minY; y < maxY; y++)
{
...
}
}
// step 2
else if (step == 2)
{
int minX = core * w / cores;
int maxX = (core + 1) * w / cores;
for(x = minX; x < maxX; x++)
<span style="white-space:pre"> </span>{
<span style="white-space:pre"> </span>...
<span style="white-space:pre"> </span>}
}
} 可以看出
分別進行了模糊水平與垂直。
同時，在模糊時：分為了許多部分，具體取決于虛擬核數。
在這里有必要說明一點就是；在模糊時只能先模糊第一步，而后才能模糊第二部。
意思就是說第一步與第二步是不能同時進行的；如果不贊同；可以啊把上面的兩個列表改成一個同時進行多線程你就知道模糊的效果了。
測試
這里截圖就不進行截圖了，不過根據我的測試；該模糊方法在多線程下優勢較為明顯；也就是說手機核數越高其模糊速度越快。
如果該方法順序執行，不使用多線程；那么其模糊所耗費時間約比我上一章中介紹的方法多1/4的時間。
下面其實，該開源項目做的很完善，同時包含了 Java，JNI C，RenderScript 三種方式進行模糊：
由于代碼較多就不一一貼出來了。我打包上傳了；需要的朋友可以自行下載。
源碼
[Android] 圖片JNI（C++\Java）高斯模糊 多線程 源碼
========================================================
作者：qiujuer
博客：blog.csdn.net/qiujuer
網站：www.qiujuer.net
開源庫：Genius-Android
轉載請注明出處：http://blog.csdn.net/qiujuer/article/details/41777001
========================================================
本文由用戶 podc9bw2 自行上傳分享，僅供網友學習交流。所有權歸原作者，若您的權利被侵害，請聯系管理員。
轉載本站原創文章，請注明出處，并保留原始鏈接、圖片水印。
本站是一個以用戶分享為主的開源技術平臺，歡迎各類分享！