Android OpenGL ES从入门到进阶（六）—— OpenGL ES人像美白与磨皮初探

于 2019-07-16 00:47:07 发布
阅读量3.1k 收藏 11
点赞数 2
分类专栏： Android OpenGL_ES
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.youkuaiyun.com/liuderong0/article/details/95936807
版权
项目源码：https://github.com/smzhldr/AGLFramework

一、前言

前面的章节中的内容差不多能让初学者对OpenGL 有个初步的理解，基础知识很重要但基础不是核心，我们该有新的目标了，基础部分忘了的可以复习下前几章的内容，想拓宽基础的可以参考文档或者看巨著。

这一节我们从OpenGL的实际应用出发，逐步来掌握OpenGL各种各样的功能，之后的的章节将以应用为背景，以进阶为目的。我们选取类似于美颜相机中的美白和磨皮两个比较典型的功能，来抛砖引玉，进一步提高。

二、效果一览

子怡姐姐得罪了，阿弥陀佛，不过美女怎么看都很美，随便调了个程度，看下效果对比还是挺明显的

三、美白浅析

在OpenGL中一个像素可以拆分成RGBA格式来处理，RGBA分别为，红、绿，蓝，透明度通道，大部分的特效处理都是按一定的规则改变不同位置像素的RGBA的值，取值为0.0 - 1.0，我们只是演示美白功能，为了简单，同时改变每个像素的RGB的值，若果RGB都为0则是黑色，RGB全为1则是白色，所以我们对每个像素做如下操作：


			vec4 color = texture2D(inputImageTexture,textureCoordinate);			
			...
			//其中0.25是取得经验值，可以随便改多试试，whiteLevel可动态调节美白程度
 			color.r= color.r + 0.25 * whiteLevel;
            color.g=  color.g + 0.25 * whiteLevel;
            color.b= color.b + 0.25 * whiteLevel；
            gl_FragColor = color;

完整的Filter长相如下：


public class WhiteFilter extends AGLFilter {

    private static final String WIHITE_FRAGMENT_SHADER =
            "precision mediump float;\n" +
                    "uniform sampler2D inputImageTexture;\n" +
                    "uniform float level;\n" +
                    "varying vec2 textureCoordinate;\n" +
                    "\n " +
                    "void modifyColor(vec4 color){\n" +
                    "color.r=max(min(color.r,1.0),0.0);\n" +
                    "color.g=max(min(color.g,1.0),0.0);\n" +
                    "color.b=max(min(color.b,1.0),0.0);\n" +
                    "color.a=max(min(color.a,1.0),0.0);\n" +
                    " }\n" +
                    "\n " +
                    "void main() {\n" +
                    "vec4 nColor = texture2D(inputImageTexture,textureCoordinate);\n" +
                    "vec4 deltaColor = nColor+vec4(vec3(level * 0.25),0.0);\n" +
                    "modifyColor(deltaColor);\n" +
                    "gl_FragColor = deltaColor;\n" +
                    "}";


    private int glUniformLevel;

    private float level;

    @Override
    protected void onInit() {
        programId = OpenGlUtils.loadProgram(NO_FILTER_VERTEX_SHADER, WIHITE_FRAGMENT_SHADER);
        glAttrPosition = GLES20.glGetAttribLocation(programId, "position");
        glAttrTextureCoordinate = GLES20.glGetAttribLocation(programId, "inputTextureCoordinate");
        glUniformTexture = GLES20.glGetUniformLocation(programId, "inputImageTexture");
        glUniformLevel = GLES20.glGetUniformLocation(programId, "level");
    }

    @Override
    protected void onDrawArraysPre(Frame frame) {
        GLES20.glUniform1f(glUniformLevel, level);
    }

    public void setWhiteLevel(float level) {
        this.level = level;
    }
}

具体使用可以看源码：https://github.com/smzhldr/AGLFramework

四、磨皮浅析

磨皮相对美白稍微要复杂一些，人物的正常肤，色应该是偏红，所以则会导致人物图像的红色通道偏亮，使红色通道保留的细节成分较少，相比之下，绿色通道保留更多细节，所以，一般情况下我们也可以只采用过滤绿色通道来实现实时磨皮。

1.取出绿色通道，对绿色通道进行模糊处理，例如高斯模糊，得到模糊后的值sampleColor
2.用原图绿色通道值减去sampleColor，加上0.5（即128），1+2两个步骤即PS中的高反差保留
3.对上述结果值进行3-5次强光处理，此步骤可以使得噪声更加突出
4.计算原图的灰度值，公式为0.299R + 0.587G + 0.114*B
5.将灰度值作为阈值，用来排除非皮肤部分，根据灰度值计算，将原图与1-3后的结果图合成
6.对混合后结果增加亮度
7.以灰度值作为透明度将原图与混合后结果进行滤色、柔光等混合，并调节饱和度

在有了美白的基础之后是不是觉得磨皮像在做数学题一样，写出公式即可，虽然步骤多一点，但是却有迹可循，上面提到的模糊一般的做法是在一个像素周围取一圈像素，然后按一定的权重取均值，混合利用OpenGL内置函数mix()即可,柔光等光线调节则与前面的美白异曲同工。

完整的Filter长这样：

public class SmoothFilter extends AGLFilter {

    private static final String SMOOTH_FRAGMENT_SHADER = "precision mediump float;\n" +
            "\n" +
            "varying mediump vec2 textureCoordinate;\n" +
            "\n" +
            "uniform sampler2D inputImageTexture;\n" +
            "uniform vec2 singleStepOffset;\n" +
            "uniform mediump float params;\n" +
            "\n" +
            "const highp vec3 W = vec3(0.299,0.587,0.114);\n" +
            "vec2 blurCoordinates[20];\n" +
            "\n" +
            "float hardLight(float color)\n" +
            "{\n" +
            "    if(color <= 0.5)\n" +
            "        color = color * color * 2.0;\n" +
            "    else\n" +
            "        color = 1.0 - ((1.0 - color)*(1.0 - color) * 2.0);\n" +
            "    return color;\n" +
            "}\n" +
            "\n" +
            "void main(){\n" +
            "\n" +
            "    vec3 centralColor = texture2D(inputImageTexture, textureCoordinate).rgb;\n" +
            "    blurCoordinates[0] = textureCoordinate.xy + singleStepOffset * vec2(0.0, -10.0);\n" +
            "    blurCoordinates[1] = textureCoordinate.xy + singleStepOffset * vec2(0.0, 10.0);\n" +
            "    blurCoordinates[2] = textureCoordinate.xy + singleStepOffset * vec2(-10.0, 0.0);\n" +
            "    blurCoordinates[3] = textureCoordinate.xy + singleStepOffset * vec2(10.0, 0.0);\n" +
            "    blurCoordinates[4] = textureCoordinate.xy + singleStepOffset * vec2(5.0, -8.0);\n" +
            "    blurCoordinates[5] = textureCoordinate.xy + singleStepOffset * vec2(5.0, 8.0);\n" +
            "    blurCoordinates[6] = textureCoordinate.xy + singleStepOffset * vec2(-5.0, 8.0);\n" +
            "    blurCoordinates[7] = textureCoordinate.xy + singleStepOffset * vec2(-5.0, -8.0);\n" +
            "    blurCoordinates[8] = textureCoordinate.xy + singleStepOffset * vec2(8.0, -5.0);\n" +
            "    blurCoordinates[9] = textureCoordinate.xy + singleStepOffset * vec2(8.0, 5.0);\n" +
            "    blurCoordinates[10] = textureCoordinate.xy + singleStepOffset * vec2(-8.0, 5.0);\n" +
            "    blurCoordinates[11] = textureCoordinate.xy + singleStepOffset * vec2(-8.0, -5.0);\n" +
            "    blurCoordinates[12] = textureCoordinate.xy + singleStepOffset * vec2(0.0, -6.0);\n" +
            "    blurCoordinates[13] = textureCoordinate.xy + singleStepOffset * vec2(0.0, 6.0);\n" +
            "    blurCoordinates[14] = textureCoordinate.xy + singleStepOffset * vec2(6.0, 0.0);\n" +
            "    blurCoordinates[15] = textureCoordinate.xy + singleStepOffset * vec2(-6.0, 0.0);\n" +
            "    blurCoordinates[16] = textureCoordinate.xy + singleStepOffset * vec2(-4.0, -4.0);\n" +
            "    blurCoordinates[17] = textureCoordinate.xy + singleStepOffset * vec2(-4.0, 4.0);\n" +
            "    blurCoordinates[18] = textureCoordinate.xy + singleStepOffset * vec2(4.0, -4.0);\n" +
            "    blurCoordinates[19] = textureCoordinate.xy + singleStepOffset * vec2(4.0, 4.0);\n" +
            "\n" +
            "    float sampleColor = centralColor.g * 20.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[0]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[1]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[2]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[3]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[4]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[5]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[6]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[7]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[8]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[9]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[10]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[11]).g;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[12]).g * 2.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[13]).g * 2.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[14]).g * 2.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[15]).g * 2.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[16]).g * 2.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[17]).g * 2.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[18]).g * 2.0;\n" +
            "    sampleColor += texture2D(inputImageTexture, blurCoordinates[19]).g * 2.0;\n" +
            "\n" +
            "    sampleColor = sampleColor / 48.0;\n" +
            "\n" +
            "    float highPass = centralColor.g - sampleColor + 0.5;\n" +
            "\n" +
            "    for(int i = 0; i < 5;i++)\n" +
            "    {\n" +
            "        highPass = hardLight(highPass);\n" +
            "    }\n" +
            "    float luminance = dot(centralColor, W);\n" +
            "\n" +
            "    float alpha = pow(luminance, params);\n" +
            "\n" +
            "    vec3 smoothColor = centralColor + (centralColor-vec3(highPass))*alpha*0.1;\n" +
            "\n" +
            "    gl_FragColor = vec4(mix(smoothColor.rgb, max(smoothColor, centralColor), alpha), 1.0);\n" +
            "}";


    private int glUniformsample;
    private int glUniformLevel;

    private float smoothLevel;

    @Override
    protected void onInit() {
        programId = OpenGlUtils.loadProgram(NO_FILTER_VERTEX_SHADER, SMOOTH_FRAGMENT_SHADER);
        glAttrPosition = GLES20.glGetAttribLocation(programId, "position");
        glAttrTextureCoordinate = GLES20.glGetAttribLocation(programId, "inputTextureCoordinate");
        glUniformTexture = GLES20.glGetUniformLocation(programId, "inputImageTexture");
        glUniformLevel = GLES20.glGetUniformLocation(programId, "params");
        glUniformsample = GLES20.glGetUniformLocation(programId, "singleStepOffset");
    }
    @Override
    protected void onDrawArraysPre(Frame frame) {
        GLES20.glUniform1f(glUniformLevel, smoothLevel);
        GLES20.glUniform2fv(glUniformsample, 1, new float[]{2.0f / frame.getTextureWidth(), 2.0f / frame.getTextureHeight()}, 0);
    }


    public void setSmoothLevel(float level) {
        this.smoothLevel = level;
    }
}

其中大部分都是shader代码，如果将shader放到文件中，filter的内容则相当简单，掌握了上面简易美白和磨皮，就可以尝试写更复杂的shader，随心所欲添加一些让人眼花缭乱的特效了，多尝试会有意想不到的效果。

具体使用可以参考
源码：https://github.com/smzhldr/AGLFramework

评论 6
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包

打赏作者

[](){}

你的鼓励将是我创作的最大动力

¥1 ¥2 ¥4 ¥6 ¥10 ¥20
扫码支付：¥1
获取中
扫码支付

您的余额不足，请更换扫码支付或充值

打赏作者

实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值