CoreImage坐标系 [英] CoreImage coordinate system

问题描述

我从 AVAsset 中获得了 CVPixelBufferRef 。我正在尝试对其应用 CIFilter 。我使用以下行：

  CVPixelBufferRef pixelBuffer = ... 
 CVPixelBufferRef newPixelBuffer = //要填充$的空像素缓冲区b $ b CIContex * context = //从EAGLContext创建的CIContext 
 CGAffineTransform preferredTransform = // AVAsset跟踪首选变换
 CIImage * phase1 = [CIImage imageWithCVPixelBuffer：pixelBuffer]; 
 CIImage * phase2 = [phase1 imageByApplyingTransform：preferredTransform]; 
 CIImage * phase3 = [self applyFiltersToImage：phase2]; 
 
 [上下文渲染：phase3到CVPixelBuffer：newPixelBuffer范围：phase3.extent colorSpace：CGColorSpaceCreateDeviceRGB（）]; 
  

不幸的是，我得到的结果方向不正确。例如，以人像模式拍摄的视频颠倒了。我想问题出在从 AVAsset 到 CoreImage 坐标系（在Xem中显示 phase2的预览） 也会显示错误的结果）。如何解决它？

  var preferredTransform = inst.preferredTransform 
 preferredTransform.b * = -1 
 preferredTransform.c * = -1 
 
 var outputImage = CIImage（cvPixelBuffer：videoFrameBuffer）
 .applying（preferredTransform）
 outputImage = outputImage.applying（CGAffineTransform（translationX：-outputImage.extent.origin.x，y：- outputImage.extent.origin.y））
  

I have CVPixelBufferRef from an AVAsset. I'm trying to apply a CIFilter to it. I use these lines:

CVPixelBufferRef pixelBuffer = ...
CVPixelBufferRef newPixelBuffer = // empty pixel buffer to fill
CIContex *context = // CIContext created from EAGLContext
CGAffineTransform preferredTransform = // AVAsset track preferred transform
CIImage *phase1 = [CIImage imageWithCVPixelBuffer:pixelBuffer];
CIImage *phase2 = [phase1 imageByApplyingTransform:preferredTransform];
CIImage *phase3 = [self applyFiltersToImage:phase2];

[context render:phase3 toCVPixelBuffer:newPixelBuffer bounds:phase3.extent colorSpace:CGColorSpaceCreateDeviceRGB()];

Unfortunately, the result I get have an incorrect orientation. For example, a video captured in the portrait mode is upside down. I guess the problem is in going from AVAsset to CoreImage coordinate system (showing a preview in XCode for phase2 also presents an incorrect result). How to fix it?

解决方案

I solved it by doing this, It should orientate everything correctly to the coordinate space

var preferredTransform = inst.preferredTransform
preferredTransform.b *= -1
preferredTransform.c *= -1

var outputImage = CIImage(cvPixelBuffer: videoFrameBuffer)
                    .applying(preferredTransform)
outputImage = outputImage.applying(CGAffineTransform(translationX: -outputImage.extent.origin.x, y: -outputImage.extent.origin.y))

这篇关于CoreImage坐标系的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！