In this paper, we present a energy consumption estimation method for microcontrollers. Vast amount of work in this topic has been done based on measuring the energy per instruction in low level programming languages like assembler. However, in our approach we use solely C programs and instructions without any regard to lower level languages. The method is based on measuring energy per each C instruction in the benchmark program and using the result to estimate the total energy consumed by the program. As a test case we use simple image processing algorithm including Gaussian blur and edge detection on PIC PIC32MX460F512L microcontroller. We show that the method is scalable on different microcontroller voltages and clock frequencies with only one set of measurements to reduce the amount of preparation work for estimation. We show that the estimation results give us credibility with less than 7% error.
