Most of digital image and signal processing algorithms are implemented into fixed-point architectures to satisfy the cost and power consumption constraints associated with embedded systems. The fixed-point conversion process is an issue for the reduction of the time-to-market and tools to automate this phase and to explore the design space are required. The ID.Fix tool based on the compiler infrastructure GECOS allows converting a C source code into a fixed-point C code using ac_fixed data types. The data word-lengths are optimized by minimizing the implementation cost under accuracy constraint. To obtain low optimization time, an analytical approach is used to evaluate the fixed-point computation accuracy. This approach is valid for systems made-up of smooth arithmetic operations.