Using Stacks for Image Segmentation Based on Region Growing

Abstract

The aim of the work is to comparatively evaluate the sizes of FIFO and LIFO stacks required for image segmentation based on growing regions. The coordinates (y, x) of the pixels that need to be attached to the cultivated area are placed in stacks during the segmentation process. The size of the stack needed to store the coordinates depends on the structure of the image and cannot be determined before segmentation. To avoid stack overflow, its size is determined for maximum load conditions when the image contains a single maximum area. In this case, the stack size is equal to the image size. This approach does not take into account the process of stack unloading and leads to memory overrun. Expressions are obtained in the paper that allow one to increase the accuracy of determining the sizes of FIFO and LIFO stacks necessary for storing the coordinates of adjacent pixels in a segmentation algorithm based on growing regions under maximum load conditions. In this case, the initial position of the region growth point and the direction of the selection of adjacent pixels in the scanning window are taken into account. A comparative assessment of the stack sizes required for image segmentation showed that using the FIFO stack is preferable to the LIFO stack and leads to significant memory savings.