The brain tumors are increasing rapidly among the younger generation. The survival of the subject can gradually be increased if the tumors are detected at early stages. Magnetic Resonance Imaging (MRI) is an important technique in detecting the tumors. The images are corrupted by random unwanted information, complicating the automatic feature extraction and the analysis of clinical data. Many methods are existing in present day to remove the unwanted information from the images. Automatic classification is essential because it reduces the cause of human error and where the accuracy is not affected. The work emphasizes on removal of noises from the MRI using the hybrid KSL technique which is the combination of Kernel, Sobel and low pass filter. Features are the properties which describe the whole image. Features from these images are extracted using shape, texture and intensity based techniques. The feature extracted are HOG and GLDM.
Recognition of vehicles has always been a desired technology for curbing the crimes done with the help of vehicles. Number imprinted on plates of cars and motorbikes are consist of numerals and alphabets, and these plates can be easily recognized. The uniqueness of combination of characters and numbers can be easily utilized for multiple purposes. For instance, fines can be imposed on people automatically for wrong parking, toll fee can be automatically collected just by recognizing the number plate, apart from these two there may be several numbers of uses can be accommodated. Computer vision is comprehended as a sub space of the computerized reasoning furthermore software engineering fields. Alternate ranges most firmly identified with computer vision are picture handling, picture examination and machine vision. As an exploratory order, computer vision is apprehensive with the counterfeit frameworks that concentrate data from pictures and recordings. The picture information can take numerous structures, for instance, segmentations of videos, taken from several cameras. This thesis presents a training based approach for the recognition of vehicle number plate. The whole process has been divided into three stages i.e. capturing the image, plate localization and recognition of digits over the plate. The characteristics of HOG have been utilized for training and SVM has been used for adopted for classifying while recognizing. This algorithm has been checked for more than 100 pictures.