Automatic scale determination for adaptive windowing in Laser Speckle Imaging

Laser Speckle Imaging (LSI) is a useful technique in medical applications since it provides information about blood flow and vascular. Traditional LSI uses standard window sizes (5×5 or 7×7) for contrast computing without taking into account the size of the ROIs (blood vessels). The speckle noise affects the whole image, however, the blood vessel region is smoother. On the one hand, a small window preserves spatial resolution but it barely attenuates the speckle noise; on the other hand, a large window significantly reduces the noise at the expense of losing spatial resolution and small ROIs. Therefore, a window size that fits with the size of the blood vessel is the one that can deal with the noise/resolution trade-off adequately. This paper proposes an automatic methodology, based on granulometry analysis, to estimate the accurate window size for contrast computation in a speckle image. The results are compared with traditional techniques for LSI.