A Pixel-based Segmentation Algorithm of Color Images by N-level-fitting

T. Hanning and G. Pisinger (Germany)


segmentation, constant piecewise approximation, con strained segments


In this article the segmentation problem is formulated as a global optimization problem. The main idea is to mini mize the Euclidean distance of a piecewise constant func tion to a given image. Of course, this problem is well de fined only if the number of constant parts has been given previously. Also the shape of the constant parts of the fit ted function must be taken in account. The result of the approximation process is constrained to the segmentation problem by using a system of allowed sets a segments . We propose to constrain the constant parts of the to be (pixel )connected. The resulting algorithm minimizes the problem by exchanging pixels between segments iteratively.

Important Links:

Go Back