Research Of Edge-Matching Image Retrieval

The modern computer technology development, causes the massive digital image to save into possibly in the computer, at the same time used the computer to carry on the retrieval to the digital image also to become a new topic of the information retrieval engineering. The multimedia data, the image, the video and so on, contains the massive information. The former antithetical couplet symbol information carried on the retrieval the method already with difficulty to utilize in the digital image. How to deal with massive image carries on the retrieval, effectively sought one kind already to become the key point of the current multimedia engineering research based on the content multimedia data retrieval technology. The goal of Based on content image retrieval (CBIR Content-Based Image Retrieval) lies in through some specific method seeks a process satisfies the human vision characteristic of image.This article uses the image edge to carry on the description to the image characteristic. It conforms to humanity's visual characteristic. When carries on to the image edge withdraws, has utilized under wavelet transform transformation characteristic, causes the image the edge to withdraw has the good anti-jamming of noise abatement ability. At the same time also it has the complete edge maintenance characteristic. Through to wavelet transform under the image dissection characteristic research, this article proposed one kind rests on wavelet transform conversion characteristics edge characteristic expression method. It separates the high-frequency partial images edge and the low frequency partial images edge effective area and carries on the description. It's able accurately to express the latent vector of the image.The image edge took the eigenvector, its eigenvector dimension and image itself has the very big relations. Because the image eigenvector dimension is unable to determine in advance, this article uses one kind of improvement Euclidean distance to match another image characteristic vector. It uses great intensity and high consideration edge as the main characteristic match the basis image.