The accuracy of the eye tracking systems is a key indicator of data validity. Currently developed eye tracking systems can be configured to be used as remote wireless autonomous systems. In order to meet the required constraints of system's responsiveness and effectively use the available hardware, image compression techniques can be employed in order to reduce the amount of data needed to be sent via a physical transmission link. Since eye tracking systems are sensitive to input image details it is necessary to preserve as much details as possible. In this paper we are presenting results of accuracy gradation for an eye tracking system depending on the quality of the decompressed image. This image is used as input for the eye tracking system in order to find the minimal acceptable quality that leads to neglectable tracking systematic errors.