Multi-Variant Enhanced Region Growing Algorithm for Medical Image Segmentation

The speckle noise presents great threat to efficient image segmentation. In the proposed novel and robust seeded region grow segmentation technique; the range of filters is additionally imposed for the improvement in image segmentation outcome. The benefits of special masks in spatial filters are effectively utilized for image smoothening. The denoised image segmentation results for frequency domain filters are compared for modified multi-variant region growing algorithm for medical image. The seeded region growing technique presented in paper is based on manual selection of seeds and implemented for computerized topographic scan chest digital image. Four adjacent unallocated pixels forming neighborhood are added after each region expansion to increase the efficiency; however this process increases the computational complexity of the algorithm. The noisy image is filtered utilizing range of filters such as spatial filter, laplacing filter, Gaussian-laplacian filter, wiener filter, inverse filter and median filter to examine and compare improvement in region growing algorithm results.