Research On Interpolation And Retrieval Algorithm Of Image Processing

This paper focused two issues, namely the interpolation and retrieval algorithm of image processing, the specific content of our research are as follows:(1) Image interpolation has a broad background of the applications; it is broadly used in the image zoom, rotation, and other geometric operations. Presently, for the image zoom operation, mature interpolation algorithms are the nearest interpolation, bilinear interpolation, as well as spline interpolation and so on. Nearest interpolation is simple and easy to achieve, faster, but will have a jagged edge and the box effects in the zoomed image; Bilinear interpolation uses the linear average weight of several neighboring pixels around the source pixels to calculate the value of the target pixels, it has a role in smoothing the edges, but the details will have a degradation, causes a loss of important characteristics of the edge domain; Cubic spline interpolation has a high smoothness, but it requires the large amount of calculations, and is liable to fuzzy edges. The above methods all assume that it is on the basis of a linear relationship between pixels and the surrounding pixels, but it has significant changes between the textures or pixels of some images, appears with a non-linear relationship. As a result, it will make the outline and the texture of the zoomed image fuzzy using the above conventional interpolation methods, especially for the pixels which do not have the characteristics of continuous gray values, lowered the image quality. The paper is based on a thorough study on the traditional intoplation algorithms, puts forward an improved edge interpolation. Our experimental results show that our proposed algorithm is good, in comparasion with traditional methods; it not only improves PSNR, but also better retains the details of the original image and maintains a more clear edge.(2) Content-Based Image Retrieval (CBIR) technology utilizes image's visual feature information (such as color, texture, shape, etc.) for image retrieval, which combines image processing, database, information retrieval, computer vision disciplines, CBIR has become a hot research at home and abroad. Based on a in-depth study of the image color feature extractions and the matching similarities, this paper presents an effective method which divides colors image into several different fields with respective weights to stand out the main field of the whole image and sets the corresponding weights for red, green and blue components according to the main hue of the colors image. Taking consideration for retrieval velocity in addition, we put forward a storied retrieval scheme based on the RGB color model. This method not only meets the needs of color image retrieval, but also enhances the precision and increases the retrieval speed. In addition, the paper makes a good research on the image retrieval based on the texture feature using fractal theory. Finally, this article designs and implements a CBIR software system based on the proposed methods.