Admissions Open
Admissions Open
Abstract:
Medical imaging refers to capturing images of different parts of the body using various methods such as X-rays, CT-scan and MRI to aide doctors and surgeons for better diagnosis. These digital medical images usually have low resolution because of the nature of their acquisition. For diagnosing the abnormality in any body part doctor usually zoom the image. The zooming can be carried out using two different methods: replication and interpolation. In replication each pixel is replicated and then each row is replicated. In this way 4 X 4 image is zoomed to a 8 X 8 image. Thus replication increases the size of the image but it produces image having patchy look. Whereas, in interpolation, the average of two adjacent pixels along the rows is taken and placed between the two pixels. The same operation is then performed along the columns. In this way size of image is increases and also the quality of image generated is better than replication. Basically, to convert a low resolution image into high resolution, interpolation technique is used which increases the number of pixels in the digital image. The interpolation method is broadly classified into three categories namely, traditional interpolation techniques, edge based interpolation techniques and wavelet based interpolation techniques. The traditional interpolation techniques such as nearest neighbor, bilinear, bicubic etc. have low processing time but the quality of interpolated image is not good. Edge based interpolation employed the concept of rendering and correction. This technique includes EDI, NEDI, iNEDI and ICBI. The visual quality of these techniques is better than the traditional techniques but the processing time is very high due to large number of iterations. Image is decomposed into four subgroups such as low-low, low-high, high-low and high-high in wavelet based techniques. In wavelet, the high frequency subbands are interpolated because edges are high frequency components of the image. These interpolated high frequency subbands are combined with the interpolated original image and then inverse discrete wavelet transform is applied on them to obtain the high resolution image. These techniques also provide the good visual results but the processing time is higher than the traditional techniques. Therefore, the research through this thesis is focused on interpolating the low resolution medical images to improve their quality and thereby getting better extent of information iv for diagnosis purposes at low processing time. So, a novel algorithm is proposed for the above said objective. The proposed algorithm is based on the point and neighborhood processing along with the gradient features of the image. The results obtained by the proposed algorithm outperforms over the existing techniques discussed in this thesis.