Image Segmentation Analysis

2998 Words12 Pages
Register to read the introduction…In image segmentation thresholding becomes a main part, it transforms a gray level image into a binary image. For that purpose some popular methods in thresholding are OTSU’s method and K- means algorithm. In my project I use OTSU method, but quite different from the standard OTSU method. Standard OTSU method was named after the scientist NOBUYUKI OTSU. It searches the histogram of an image and finds the threshold value from its binary image and segments the image into two parts foreground and background. But if we use the standard method for image segmentation there will be a defect in the image quality, because while segmenting the image into two parts that is foreground and background we may lose some pixel points . maybe the points are very close to the threshold value but since its lower than the constant value we have to ignore that point, and therefore that may be an intergral part in the image . so this becomes a disadvantage . so in my project while doing OTSU method we divide the image into three parts that is foreground, background and TBD( to be determined) region. And in the succeeding process the To Be Determined region is again divided into the same three regions again, which forms iterations . Till the threshold value becomes lower the partition continues and at last the foreground and background regions are summed together and we get the segmented image with high quality…show more content…
Since all the analysis process was always done in a binary image every image if it is either color or gray scale it has to be converted to a binary image. This is what done in otsu method. To create a binary image thresholding plays a very important role, after taking the threshold value from the pixels using the generative formula each pixel value are been compared with the threshold value and changed to either 1 or 0. So basically it was said as a clustering-based process. And in this method it converts the binarized image into two segments known as foreground and background, where foreground deals with the 1’s and background deals with the 0’s. But this method didnot prove to be a great success in image segmentation. Eventhough it gave a new way to thresholding it had a lots of limitations too. The variance always remains small or minimum just because the two segments squared distance were constant and leading to a high inter-class variance. And otsu method only dealt with one dimensional images. These reasons lead to more new methods but keeping the otsu method as a base

More about Image Segmentation Analysis

Open Document