This well-structured paper presents a method for moving object detection based on the logarithm of the intensity of the image. The text has a strong mathematical basis, and a large set of bibliographic references.
I did find several deficiencies in the text. First, only one technique (background subtraction) is assumed to exist for extracting moving objects. In fact, others do exist, and some (for example, temporal difference) are well known and widely used. Second, several approaches to logarithmic image processing are described, however the authors decide to use homomorphic filtering, without explaining why other methods are refused. Logarithmic image processing (LIP) has been found [1] to behave better than the other models.
The experimental results section seems to have been reduced to analyzing just one homemade video, with no references to it. Thresholded images are compared with different threshold measures, depending on the intensity domain (60, 70, and 80 for the standard-intensity domain, versus 0.6, 0.7, and 0.8 for the logarithmic-intensity domain). The text does not explain the reason for that change of range. The authors also refer to “desired motion pixels,” however, there is not a clue given on how the authors decide whether or not a pixel is desired.
On the positive side, this paper presents a new and interesting way to solve a recurrent problem in moving object detection: illumination changes. This technique seems to work better than the usual intensity-domain techniques, being less sensitive to changes in threshold.