1. 技术文章
  2. [Android] 图片JNI（C++\Java）高斯模糊 多线程

[Android] 图片JNI（C++\Java）高斯模糊 多线程

由 跟風遠走 提交于 2019-12-05 18:06:37

========================================================
作者：qiujuer
博客：blog.csdn.net/qiujuer
网站：www.qiujuer.net
开源库：Genius-Android
转载请注明出处：http://blog.csdn.net/qiujuer/article/details/41777001
========================================================

在我的博客中，曾经发布了一篇高斯模糊（堆栈模糊）的文章；在其中使用了国外的一个堆栈模糊来实现对图片的模糊处理；同时弄了一个JNI C++ 的版本。

这篇文章依然是堆栈模糊；可以说最原始的地方还是堆栈模糊部分；只不过是支持多线程的。

纳尼？？感情是之前那个不支持多线程？Sorry，我说错了；两个都是支持多线程调用的。不过新讲的这个是能在内部采用多线程进行分段模糊。

原来的：[Android]-图片JNI（C++\Java）高斯模糊的实现与比较

开工吧

说明：其中代码大部分来源于网络，不过都是开源的。

最原始的代码：

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;
}
这个是C++的代码。

在网上发现了一个开源的 Android 模糊的例子，其中有该代码的 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;
		}

	}
}
该代码与C++的代码同出一源。下面来简要分析一下。

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);
	}
方法进入后，首先得到图片的像素点集合；而后 New 两个 List ，分别存储 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 是用于进行 线程池的回调类；其主要调用了方法  blurIteration进行模糊。

同时给方法，传入的参数有：

  • 像素点集合
  • 宽
  • 高
  • 模糊半径
  • 最大内核数
  • 当前内核

步骤

而后回到第一步处：

int cores = StackBlurManager.EXECUTOR_THREADS;
其实是调用的： 

Runtime.getRuntime().availableProcessors();
该方法用于返回当前虚拟机的核数；用此方法可以 最大效率的进行多线程的分配。

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，对应的线程回调操作；分别是 步骤1，和步骤2

而后：

try {
			StackBlurManager.EXECUTOR.invokeAll(horizontal);
		} catch (InterruptedException e) {
			return null;
		}

		try {
			StackBlurManager.EXECUTOR.invokeAll(vertical);
		} catch (InterruptedException e) {
			return null;
		}
该方法中： StackBlurManager.EXECUTOR 其实是: 

Executors.newFixedThreadPool(Runtime.getRuntime().availableProcessors());
可以看出其实是建立了一个线程池，线程池的运行线程数为当前虚拟机的核数。

然后把步骤1的集合丢进线程池并等待执行完成；而后丢进步骤2.

其实从其命名就能看出：

  • 步骤1其实是模糊水平方向
  • 步骤2其实是模糊垂直方向

具体是不是可以看看 blurIteration 方法。

由于方法太长就简化了一下：

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 分别进行了模糊水平与垂直。

同时，在模糊时：分为了许多部分，具体取决于虚拟核数。

在这里有必要说明一点就是；在模糊时只能先模糊第一步，而后才能模糊第二部。

意思就是说第一步与第二步是不能同时进行的；如果不赞同；可以啊把上面的两个列表改成一个同时进行多线程你就知道模糊的效果了。

测试

这里截图就不进行截图了，不过根据我的测试；该模糊方法在多线程下优势较为明显；也就是说手机核数越高其模糊速度越快。

如果该方法顺序执行，不使用多线程；那么其模糊所耗费时间约比我上一章中介绍的方法多1/4的时间。


下面其实，该开源项目做的很完善，同时包含了 Java，JNI C，RenderScript  三种方式进行模糊：



由于代码较多就不一一贴出来了。我打包上传了；需要的朋友可以自行下载。

源码

[Android] 图片JNI（C++\Java）高斯模糊 多线程 源码

========================================================
作者：qiujuer
博客：blog.csdn.net/qiujuer
网站：www.qiujuer.net
开源库：Genius-Android
转载请注明出处：http://blog.csdn.net/qiujuer/article/details/41777001
========================================================

标签
JNI
Cc++
android
高斯模糊
堆栈模糊
多线程
易学教程内所有资源均来自网络或用户发布的内容，如有违反法律规定的内容欢迎反馈！
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!