SDAA422 June 2026 TDA4VE-Q1
Texas Instruments Processor SDK RTOS provides the TIOVX and vision_apps frameworks for developing heterogeneous image processing and AI pipelines on TDA4x devices. While the SDK includes a comprehensive set of standard OpenVX kernels, many applications require custom processing functions that are not supported by the built-in kernel library. This application note describes the complete workflow for developing and integrating a custom OpenVX node that executes on an Arm Cortex-A72 core, including kernel interface definition, Host and Target Kernel implementation, kernel registration, build integration, and graph execution within the vision_apps framework. A grayscale threshold operation is used as a reference implementation to illustrate parameter validation, shared-memory buffer access, and runtime processing. Although a simple threshold algorithm is used for demonstration purposes, the methodology presented in this document can be readily applied to application-specific image processing, sensor processing, and AI pre-processing or post-processing functions.