Method for processing an image sequence and tester for a car

A method for recognizing motion in an image sequence, in which an image processing device ascertains a change of position for current image features in a current image with respect to corresponding reference image features in a reference image, the method comprising:

• identifying mutually corresponding image features by pairing current image features in the current image and reference image features in the reference image;
• ascertaining at least one steady region for the current image, the at least one steady region having current image features that have a change of position with respect to the reference image that is less than a predetermined degree;
• determining model parameters for a motion model to describe the change of position of only mutually corresponding image features that are situated outside the at least one steady region;
• transforming the reference image with the model parameters to produce a transformed reference image;
• checking the model parameters using the image processing device, the model parameters being checked to determine whether the change of position comprises a rotation;
• and
• transmitting the transformed reference image to a compression device of a video coder as a motion-compensated reference image, the transformed reference image being transmitted such that if the rotation is determined, the transformed reference image is transmitted only if an angle of rotation is greater than a predetermined threshold value.

A tester for a motor vehicle comprising:

• an image processing device to receive an image sequence from a navigation appliance in the motor vehicle and to ascertain model parameters for a motion model for a current image from the image sequence for recognizing motion;
• and
• a video coding device to code the image sequence,
• wherein the image processing device is designed to:
• ascertain a change of position for current image features in the current image with respect to corresponding reference image features in a reference image;
• identify mutually corresponding image features by pairing current image features in the current image and reference image features in the reference image;
• ascertain at least one steady region for the current image, the at least one steady region having current image features that have a change of position with respect to the reference image that is less than a predetermined degree;
• determine model parameters for a motion model to describe the change of position of only mutually corresponding image features that are situated outside the at least one steady region;
• transform the reference image with the model parameters to produce a transformed reference image;
• check the model parameters, the model parameters being checked to determine whether the change of position comprises a rotation;
• and
• transmit the transformed reference image to a compression device of a video coder as a motion-compensated reference image, the transformed reference image being transmitted such that if the rotation is determined, the transformed reference image is transmitted only if an angle of rotation is greater than a predetermined threshold value.