在我写的上篇文章中,介绍了美颜滤镜的实现原理,已经能够体会到 GPUImage 的强大。本文将要介绍的Faceu贴纸效果也是基于GPUImage实现的,demo我放在了GitHub上。
1.核心原理
Faceu贴纸效果其实就是在人脸上贴一些图片,同时这些图片是跟随着人脸的位置改变的。如果我们不强调贴图的位置,这就是一个简单的水印需求。
根据人脸检测的结果动态调整水印贴纸的位置即可实现简单的Faceu效果。
2.水印
在GPUImage的官方demo中就已经有文字水印的实现:
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 |
GPUImageFilter *filter = [[GPUImageFilter alloc] init]; [self.videoCamera addTarget:filter]; GPUImageAlphaBlendFilter *blendFilter = [[GPUImageAlphaBlendFilter alloc] init]; blendFilter.mix = 1.0; NSDate *startTime = [NSDate date]; UIView *temp = [[UIView alloc] initWithFrame:self.view.frame]; UILabel *timeLabel = [[UILabel alloc] initWithFrame:CGRectMake(0.0, 0.0, 240.0f, 40.0f)]; timeLabel.font = [UIFont systemFontOfSize:17.0f]; timeLabel.text = @"Time: 0.0 s"; timeLabel.textAlignment = UITextAlignmentCenter; timeLabel.backgroundColor = [UIColor clearColor]; timeLabel.textColor = [UIColor whiteColor]; [temp addSubview:timeLabel]; uiElementInput = [[GPUImageUIElement alloc] initWithView:temp]; [filter addTarget:blendFilter]; [uiElementInput addTarget:blendFilter]; [blendFilter addTarget:filterView]; __unsafe_unretained GPUImageUIElement *weakUIElementInput = uiElementInput; [filter setFrameProcessingCompletionBlock:^(GPUImageOutput * filter, CMTime frameTime){ timeLabel.text = [NSString stringWithFormat:@"Time: %f s", -[startTime timeIntervalSinceNow]]; [weakUIElementInput update]; }]; |
要理解它的实现原理,需要搞懂GPUImageUIElement和GPUImageAlphaBlendFilter。GPUImageUIElement的作用是把一个视图的layer通过CALayer的renderInContext:方法把layer转化为image,然后作为OpenGL的纹理传给GPUImageAlphaBlendFilter。而GPUImageAlphaBlendFilter则是一个两输入的blend filter, 它的第一个输入是摄像头数据,第二个输入则是刚刚提到的GPUImageUIElement的数据,GPUImageAlphaBlendFilter将这两个输入做alpha blend,可以简单的理解为将第二个输入叠加到第一个的上面,更多关于alpha blend在维基百科上有介绍。下图是整个加水印的过程:
3.人脸检测
利用CIDetector即可简单的实现人脸检测,首先是CIDetector的初始化:
1 2 3 |
NSDictionary *detectorOptions = [[NSDictionary alloc] initWithObjectsAndKeys:CIDetectorAccuracyLow, CIDetectorAccuracy, nil]; _faceDetector = [CIDetector detectorOfType:CIDetectorTypeFace context:nil options:detectorOptions]; |
然后通过将摄像头数据CMSampleBufferRef转化为CIImage,对CIImage用CIDetector进行人脸检测:
1 2 3 4 |
CVPixelBufferRef pixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer); CFDictionaryRef attachments = CMCopyDictionaryOfAttachments(kCFAllocatorDefault, sampleBuffer, kCMAttachmentMode_ShouldPropagate); CIImage *convertedImage = [[CIImage alloc] initWithCVPixelBuffer:pixelBuffer options:(__bridge NSDictionary *)attachments]; NSArray *features = [self.faceDetector featuresInImage:convertedImage options:imageOptions]; |
上面得到的features数组里的每个元素都是CIFaceFeature对象,根据它就能计算出人脸的具体位置,从而调整中水印图像的位置,达到图像跟随人脸动的效果。
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 |
for ( CIFaceFeature *faceFeature in featureArray) { // find the correct position for the square layer within the previewLayer // the feature box originates in the bottom left of the video frame. // (Bottom right if mirroring is turned on) //Update face bounds for iOS Coordinate System CGRect faceRect = [faceFeature bounds]; // flip preview width and height CGFloat temp = faceRect.size.width; faceRect.size.width = faceRect.size.height; faceRect.size.height = temp; temp = faceRect.origin.x; faceRect.origin.x = faceRect.origin.y; faceRect.origin.y = temp; // scale coordinates so they fit in the preview box, which may be scaled CGFloat widthScaleBy = previewBox.size.width / clap.size.height; CGFloat heightScaleBy = previewBox.size.height / clap.size.width; faceRect.size.width *= widthScaleBy; faceRect.size.height *= heightScaleBy; faceRect.origin.x *= widthScaleBy; faceRect.origin.y *= heightScaleBy; faceRect = CGRectOffset(faceRect, previewBox.origin.x, previewBox.origin.y); //mirror CGRect rect = CGRectMake(previewBox.size.width - faceRect.origin.x - faceRect.size.width, faceRect.origin.y, faceRect.size.width, faceRect.size.height); if (fabs(rect.origin.x - self.faceBounds.origin.x) > 5.0) { self.faceBounds = rect; } } |
上面则是计算人脸位置faceBounds的方法,我们再根据faceBounds来更新水印图像的位置:
1 2 3 4 5 6 7 8 9 |
__weak typeof (self) weakSelf = self; [filter setFrameProcessingCompletionBlock:^(GPUImageOutput *output, CMTime time) { __strong typeof (self) strongSelf = weakSelf; // update capImageView's frame CGRect rect = strongSelf.faceBounds; CGSize size = strongSelf.capImageView.frame.size; strongSelf.capImageView.frame = CGRectMake(rect.origin.x + (rect.size.width - size.width)/2, rect.origin.y - size.height, size.width, size.height); [strongSelf.element update]; }]; |
4.延伸
- 问题1:上面用的人脸检测是基于CIDetector的,实际实验发现,当人脸在摄像头中捕获不全时,有可能检测不出人脸,也就没法更新水印图像的位置。因此,更加精准、快速、细致的人脸检测是很有必要的,后面我会尝试使用一些其他的人脸检测方法。
- 问题2:上面的Faceu贴纸效果是静态图像的贴纸效果,如果要做动态效果的Faceu贴纸该怎么处理呢, Gif? CADisplayLink? 这个有待进一步研究,如果有这方面经验的朋友也欢迎在评论区留言,互相交流学习。