Figure 2 presents a simplified diagram of the flow that the tutorial implements:

  1. The grayscale image that the Canny requires is extracted as Y (luminance) channel from the input NV12 YUV image.
  2. U and V channels are extracted as well to further combine in output image.
  3. Canny edge detection is applied to Y channel.
  4. Modified Y, and original U and V channels are combined in output image.

Figure 2: High-level simplified data flow implemented in the OpenVX part of the tutorial.

Figure 3 presents the complete OpenVX* graph:

Figure 3: Complete OpenVX graph implemented in the tutorial.

Legend: nodes are green, data objects are blue, "virt" indicate virtual objects.

Names for the data objects correspond to the names of variables in the source code.

Para obter informações mais completas sobre otimizações do compilador, consulte nosso aviso de otimização.