Subscribe to
NSLog(); Header Image

CoreImage Transcoding

CoreImage: slow?

We're finding that as we work on Rivet the bottleneck isn't the network, but the image itself. We're performing two operations: resizing (to the larger of 1920 x 1080) and compressing (currently, the options are none, 90, 70, 50, and 30). These operations all seem to take about the same amount of time, and that time is slower than most networks can probably just push the extra 2 MB of data we're saving per image.

Is that realistic? Is CoreImage that slow?

One Response to "CoreImage Transcoding"

  1. Apparently CILanczosScaleTransform is rather slow since it produces a high quality result. It also seems to depend on the GPU hardware. Googling for "CILanczosScaleTransform slow" produces many hits. Suggested workarounds are to use CIAffineTransform if you're willing to accept a lower quality result.