Edge-preserving Wavelet Image Representations with Applications to Compression of Aerial Ortho Images

This paper describes a new second-generation-type lossy image compression scheme using wavelets for aerial ortho images. For map applications where some losses are allowed, aerial ortho images present a possibility of having a very high compression ratio, because they are natural images with many uniform (coherent) subregions. However classical compression schemes such as the joint-photograph expert group (JPEG) standard do not preserve image edges, which are critical in preserving precision of the objects' positions. In our compression scheme, an image is seen as a collection of edges and textures, instead of pixels, hence edge preservation can be made explicit. This scheme is particularly promising because the images contain many man-made, regular objects, such as roads, houses, parks, and farmlands. The scheme uses wavelet extrema to extract the edges. The paper shows that an excellent reconstruction of aerial ortho images can be obtained from edges. The paper also discusses techniques to compress edges of the aerial ortho images.