In this dissertation, a method and a tool to enable design and verification of computation demanding embedded vision-based systems is presented. Starting with an executable specification in OpenCV, we provide subsequent refinements and verification down to a system-on-chip prototype into an FPGA-Based smart camera. At each level of abstraction, properties of image processing applications are used along with structure composition to provide a generic architecture that can be automatically verified and mapped to the lower abstraction level. The result is a framework that encapsulates the computer vision library OpenCV at the highest level, integrates Accelera's System-C/TLM with UVM and QEMU-OS for virtual prototyping and verification and mapping to a lower level, the last of which is the FPGA. This will relieve hardware designers from time-consuming and error-prone manual implementations, thus allowing them to focus on other steps of the design process. We also propose a novel streaming interface, called Component Interconnect and Data Access (CIDA), for embedded video designs, along with a formal model and a component composition mechanism to cluster components in logical and operational groups that reduce resource usage and power consumption.