The importance of edges in irregular subsampling

The quality of the reconstructed image, from a very weak set of irregularly spaced pixels, depends strongly on the localization of pixels on the image grid. Two different subsampling methods show that the distribution of the pixels is closely linked to the characteristics of the image. The first method, based on the second generation of wavelets decomposes the original image in several subsets. Pixels belonging to each subset are distributed according to a threshold. The second method chooses the subset minimizing the mean square error of the reconstructed image. The reconstruction method is a two-dimensional statistical interpolation method. Both subsets of pixels, provided by the two methods, have similar characteristics. Experimental results are provided and show that most of pixels are located in sharp edge regions.