期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2013
卷号：1
期号：4
出版社：S&S Publications
摘要：In computer vision, image compression mainly refers to the problem of reducing the memory size of a digitalimage. Mainly problems regarding data transferring over internet might require an image data to be of comparativelylesser memory size. Moreover high bandwidth is also required for transmission of high quality image data. Imagecompression provides solution to this problem. Now in order to retrieve the image of lesser size there is a considerableamount of loss of image data. For this problem optimized method is chosen to provide a final image which iscomparatively of smaller memory size than the original image, yet it is quite visually similar to the original image. Inthis paper, we are trying to segment the original image using K-Means clustering method. Let us consider a grey levelimage of M XN size, now we know there are in total 0-255 grey level intensity values i.e 8 bits for each pixel thereforethe total size of the image will be M XN X8 bits. Now in our approach we have segmented the image into K clusters.Let the set of cluster centres be (c1, c2,..........., ck) i.e. we have seen that on segmentation of the image into 8 (k=8)segments better results are obtained, hence the total image can be replaced accordingly with 8 grey level intensityvalues, so the 8 values can be represented by 3 bits (000, 001, ……..,111). Now if we represent the image with 8 greylevel values there is a high loss of data. In our approach we have tried to minimize the loss of data using correlativecoefficient function. In every iteration of K-Means clustering algorithm we have compared the segmented image withthe original image until the loss of data is minimized (the value of correlative co-efficient function should bemaximum) and then the iteration is stopped. In this approach the compressed image is of M XN X3bits memory size,hence there is 62.5% decrease of the memory size of the compressed image, but still there is above 90% visualsimilarity of the compressed image with the original image. So in our approach the main bottleneck of imagecompression is satisfied considerably. Advanced techniques of image processing and analysis find widespread use inmedicine. In medical applications, image data are used to gather details regarding the process of patient imagingwhether it is a disease process or a physiological process. Information provided by medical images has become a vitalpart of today‟s patient care. The images generated in medical applications are complex and vary notably fromapplication to application. Nuclear medicine images show characteristic information about the physiological propertiesof the structures-organs. In order to have high quality medical images for reliable diagnosis, the processing of image isnecessary.
