Medical Image Retrieval Using Haralick Texture Features
Abstract
Medical imaging has been placed as one of the most significant medical developments over the past 1,000 years. On the domain of applications in computer science with respect to the field of medicine, the medical image data procedure is playing an improved specific role. The medical data provides valuable information that will be used for diagnosis, recovery, treatment, rehabilitation, etc. CT, X-Ray, MRI and ultrasound are medical imaging techniques that are the amount of images digitized and devised in hospitals are incredibly increasing. Thus, the requirement of effective retrieval system for particular image is very high. Unfortunately, only some of the medical retrieval systems for images are used in clinical side. Avoidance of risk factors is the first method of prevention. Image retrieval is the basic requirement of present life. As more images are added in database, from various resources for the image retrieval, different types of processing is required to retrieve the features from the images. The aim of medical image retrieval is to retrieve the most similar medical images in response to required information requests represented as search queries. Texture feature extraction has important applications such as medical images and remote sensing. In this project, the images are retrieved by the CBIR system. The CBIR aim is to query and find images in database and retrieve the relevant images from the database. This discussed work describe, the information present in the image database through Haralick Texture Features by GLCM, for analysing the texture. The GLCM was proposed by Haralick to get the statistical features in texture. The GLCM contains the statistical information of the second-order of spatial relationship of image pixels. We use five Haralick Texture Features for retrieving similar images. The retrieved results will be evaluated with precision and recall of the system. The proposed retrieval results can be able to detect various diseases.
