The Research On Image Mosaic Technology Based On Improved FREAK Algorithm

As an important subject in computer vision, image mosaic is widely used in many fields and daily life. Image mosaic technology is to solve the problem of both wide angle and high spatial resolution. This has become a hot spot for a large number of scientific research workers, so that ordinary imaging equipment and general digital cameras are able to obtain images of high resolution and wide field. Image mosaic technology is according to a part of overlap of the image sequence and in the case of not reducing the image sharpness using image processing algorithm on the computer, registration of image characteristic information,and then use the fusion algorithm to make the stitched image has characteristics of wide field of vision, high definition, seamless, and so on. That is to say, the process of stitching is mainly divided into two important steps: image registration and image fusion.This paper first introduces the basic principles and related technology of image stitching.For image registration elaborated several traditional feature points matching algorithm. After comparing we found that the disadvantages image matching in different scales with the traditional method,and put forward a feature point matching algorithm based on improved FREAK algorithm. Improved algorithm combine SURF algorithm with FREAK algorithm for image matching. Firstly, using Hessian matrix of SURF algorithm to determine the candidate points for suppression of the non maximum, in order to establish scale space. Then the feature points were described by the FREAK descriptors to distribute directions for the feature points.Finally, the image matching was completed. The experimental results show that the feature points matching accuracy of the improved algorithm is improved, there is a good matching effect on the scale difference, light intensity difference and the rotation difference. According to a part of overlap of two original images, using the improved FREAK feature point matching algorithm that is conducive to image alignment accuracy combines RANSAC algorithm to purify the matching and establish a single stress matrix to unify the two images in the new coordinate system. A linear gradient weighted algorithm is used to eliminate the registration of the joint to achieve the image mosaic technology. Finally, a width of vision,high-quality mosaic stitched image is obtained. It has good visual effect.According to image fusion an improved image fusion algorithm based on wavelet transform is proposed. Wavelet transform decomposes the image into high frequency coefficients and low frequency coefficients. The high frequency coefficients are processed by using convolution of two kinds of Laplacian template operator, and the image entropy function is used as the evaluation and judgment function. And then the final numerical value is obtained from the comparison of the entropy function. The low frequency coefficients with Otsu algorithm are processed by using the adaptive judgment method of the threshold and Laplacian definition evaluation function. This fusion rule limit the maximum energy of the original image extraction, as far as possible to retain the original image of the image. Finally,the fused image is obtained by inverse wavelet transform. The effectiveness of the improved algorithm is illustrated by subjective image quality evaluation and objective image quality evaluation.