3.2.1. Image Pre-Processing

In the identification and verification of blood flow at the upper part of the hand, the first stage is to locate the point of interest (ROI). It is identified as the part of image that has the object of interest. In this context, the object of interest is the bloodstream at the upper part of the hand (this is based on a pre-selected biometric characteristic). To define the location and the value of ROI, the position and the shape of the hand needs to be identified. Extracting blood stream biometric characteristics implies gathering information on the biometric characteristics of the hand. Figure 4 shows how ROI is derived (green rectangle). The ROI area starts from the joints level, and is associated with the vector that exists from finger valley point numbers 2 and 4, and up to the middle of the hand. This helps to differentiate (through ROI width) each user and/or specific individual image. The height of ROI is determined and expressed as a multiple value (precisely in multiples of 1.4) of the width.

**Figure 4.** Process of bloodstream filtration: (**a**) before filtration (left), (**b**) after median filtration (middle), and (**c**) after equalization (right).

During the second phase, the ROI is used to develop a copy of the new image, which is subjected to rotation. The goal of image rotation is to ensure that the orientation of the bloodstream image is vertical and corresponds with the longer side. Figure 4 shows the result after copying ROI (Region of Interest) and rotating the image.

Rotation and equalization are followed by image filtering. In the context of the proposed system, the image is filtered using a median filter with a relatively large surrounding of the points (11 pixels). As such, significant image smoothening is achieved, so that skin wrinkles and body hair distortions are removed. However, the shape of the veins is retained on the image (Figure 4b).

To further treat the images, equalization of the histogram was carried out. This process adjusts the brightness of individual pixels relative to the histogram. For the images with a similar intensity of pixels, as in this case, the contrast becomes better. Using this procedure, the difference between the background and the object of interest (bloodstream) increases (Figure 4c).
