Adaptive coding unit (CU) partitioning based on image statistics
A method for determining coding unit (CU) partitioning of a largest coding unit (LCU) of a picture is provided that includes computing a first statistical measure and a second statistical measure for the LCU, selecting the LCU as the CU partitioning when the first statistical measure does not exceed a first threshold and the second statistical measure does not exceed a second threshold, and selecting CUs in one or more lower layers of a CU hierarchy of the LCU to form the CU partitioning when the first statistical measure exceeds the first threshold and/or the second statistical measure exceeds the second threshold.
1. A method comprising:
receiving, by a processor, a largest coding unit (LCU) of a video;
determining, by the processor, a first variance and a first gradient of the LCU;
responsive to the first variance not exceeding a first variance threshold and the first gradient not exceeding a first gradient threshold, selecting, by the processor, the LCU as a partition; and
responsive to either or both the first variance exceeding the first variance threshold and the first gradient exceeding the first gradient threshold:
partitioning, by the processor, the LCU into one or more respective layers of CUs, wherein each of the one or more respective layers of CUs is associated with a respective variance threshold and a respective gradient threshold;
determining, by the processor, a variance and a gradient for each CU of each of the one or more respective layers of CUs;
comparing, by the processor, the variance for each CU of the respective layer of CUs with the respective variance threshold associated with the respective layer of CUs;
comparing, by the processor, the gradient for each CU of the respective layer of CUs with the respective gradient threshold associated with the respective layer of CUs; and
for each CU of the respective layer of CUs, responsive to the variance not exceeding the respective variance threshold associated with the respective layer and the gradient not exceeding the respective gradient threshold associated with the respective layer of CUs, selecting, by the processor, the CU as a partition for the LCU,
wherein the respective variance threshold and the respective gradient threshold for each of the respective layers of CUs are each based on an expected content of the video.
2. The method of claim 1 , wherein computing the variance for each CU of the respective layers of CUs comprises using a sum of a squared difference between each pixel in each of the respective layers of CUs and the average pixel value in each of the respective layers of CUs.
3. The method of claim 1 , wherein computing the gradient for each CU of the respective layers of CUs includes:
computing a horizontal gradient of the pixel values of each of the respective layers of CUs; and
computing a vertical gradient of the pixel values of each of the respective layers of CUs.
4. The method of claim 1 , wherein:
the respective variance threshold and the respective gradient threshold for each of the respective layers of CUs are each based on an application.