Comparison of the Work Process Methods for Scale Invariant Feature Transform and Sum of Squared Difference in Image Mosaic

Abstract

Mosaic image is an image made from a collection of many other small images placed side by side, so that from a distance it looks like one big image. The use of this mosaic image varies, for example in digital image processing, in the context of image analysis, mosaic can also mean the process of combining several images with overlap to create one large image, for example in geographic mapping using satellite imagery. Then as a pattern creation in the field of design, mosaic images are used to create complex or abstract patterns, which can be used in decoration, architecture, or product design. Mosaic image is a combination of several images to get a wider view. One of the problems in mosaic image is in the image matching process, the right image matching can produce a better mosaic image. This study will compare the image matching method with RANSAC and SSD. The mosaic image of both methods was tested using objective fidelity criteria. The results showed the RANSAC method with an MSE value of 121.5820 and a PSNR value of 27.2821 dB, while the SSD method with an MSE value of 140.8373 and a PSNR value of 26.6436 db. The RANSAC algorithm is good for use in cases of mosaic images with feature-based methods, while SSD is good for use in cases of mosaic images with direct methods.