Pathological Motion

author: David Corrigan
keywords: Motion Estimation, Motion Estimation Failure, Missing Data Treatment, Missing Data Detection

Pathological Motion (PM) is object motion which cannot be estimated reliably by standard motion estimators. PM is often associated with objects which are being occluded in a scene or with fast moving objects (often causing motion blur). Although not a form of degradation itself, the presence of PM can cause restoration applications to fail. The group has been concerned with the effect that PM has on missing data treatment.

Pathological Motion and Missing Data Treatment

PM reduces the robustness of Missing Data Treatment algorithms (ie "dirt and sparkle" or blotch removal) as it causes regions of real image data to be mistaken as blotches. As a result, it is common to see these regions removed from images after treatment (Fig. 1 & 2). The adopted solution to this problem [1], is to use more frames in which to detect blotches (up to 5 from the standard 3). This makes it easier to distinguish between actual blotches and false alarms caused by PM.


Figure 1: A sequence with PM. In this sequence the propellor blades are appearing and disappearing repetitively.
Figure 2: This sequence is a restored sequence using a standard missing data detector. Because of the Pathological Motion of the blades, they are often detected as regions of missing data when a standard missing data detector is used.
Figure 3: This video shows the sequence restored using the PM robust missing data detector. PM is no longer detected as missing data and the blades remain intact.
Figure 4: The final clip shows the output of a standard blotch detector (top) and the PM robust detector (bottom) superimposed on a section of each frame. The green colour represents detected missing data and the red colour represents detected PM. This shows that the motion of the blades and airmen causes false detections and that the PM robust detector detects these regions as PM.


  1. Pathological Motion Detection for Robust Missing Data Treatment in Degraded Archived Media. D. Corrigan , N. Harte and A. Kokaram (2006) In IEEE ICIP. Atlanta, USA, pages 621-624.
Page last modified on September 28, 2007