Complex 2D matched filtering without Halo artifacts

Accurate extraction of previously defined objects from noisy images is still a challenge for most image processing systems. There is a trade-off between the unwanted and useful amount of extracted information. From quality aspect, the matched filtering approach is preferable, as it significantly reduces the amount of unwanted information; however, the segmentation of filtered images usually is difficult. Complex matched filtering does the opposite: it provides the additional information, which simplifies the segmentation procedure; however, the resulting images are filled with the unwanted artifacts. This paper introduces the improved 2D complex matched filtering approach. The approach discussed in this paper doesn´t produce the unwanted artifacts and still retains the angular information about the extracted objects. This improvement is significant for the systems that are sensitive to noise and demand precise segmentation results.