The aim of this master thesis was to develop a solution for determining the distance of the detected vehicle based on a stereo pair of images and implement it on a realistic ADAS platform. The complete software solution for this thesis was performed in the Vision SDK software platform and was implemented on the ADAS ALPHA development board. The proposed final solution to the thesis’s task is based on the application of FGS filter with guidance over the calculated disparity map, in order to smooth the surfaces representing certain objects of the captured 3D scene, while preserving the edges of the objects. The test results of the final solution are compared with the results of the solution for determining the distance of the detected vehicle based on the usage of images obtained from one camera and with the results of the solution for determining the distance of the detected vehicle based on the usage of stereo image pairs, without FGS filter. Four different vehicles were used for testing, for which the distance was estimated at six test positions. Using the test results, it is possible to see that a solution based on the usage of an image obtained from a single camera gives results with the least relative error and is performed the fastest. Also, it is clearly seen that a solution based on the use of stereo image pairs, with the application of FGS guided filter, results in values with less relative error than the values obtained by a solution based on the usage of stereo image pairs, but without the application of FGS filter. From the above, it can be concluded that the usage of FGS filter with guidance makes sense. Possible improvements to the shortcomings of the proposed final solution, such as faster execution and less distance estimation error, can be implemented in the form of improvements to the vehicle detection algorithm, the calibration algorithm, the algorithm for correcting stereo image pairs and the optimization of the execution time.