Image surface approximation with irregular samples

A method to approximate image surfaces using irregular samples is proposed. Experimental results show that a high compression ratio can be achieved for simple or complicated images. A method accounting for the three-dimensional effects is proposed as a way of selecting the irregular samples from the image. Approximation spline knots are chosen along the contours, and the knot replacement is sensitive to the curvature along contours and the density of contours in the near neighborhood. In general, the proposed method was shown to reduce the reconstruction error both qualitatively and quantitatively, and can be used to model the image surface for the purpose of data compression and representation. This approach retains key features in images and throws away redundant data.<
>