Image segmentation using a hierarchical sequence of piecewise constant approximations that minimally differ from the original image in terms of the total squared error is discussed. It is proposed to
В статье анализируется проблема сегментации цветового изображения, аппроксимируемого кусочно-постоянными приближениями. Качество сегментации оценивается по классическому среднеквадратичному отклонению (СКО) пикселей приближения от пикселей изображения. Обсуждаются современные версии классических методов кластеризации пикселей изображения посредством минимизации СКО или суммарной квадратичной ошибки. Описываются четыре основные операции с кластерами пикселей и критерии их выполнения для построения оптимизированных приближений. Предлагаются варианты алгоритма преобразования приближения изображения, которые при неизменном числе сегментов обеспечивают оптимизацию приближения как по СКО, так и по зрительному восприятию.
18 September 2013 in Pattern Recognition and Image Analysis
The Otsu method (1979) and the Mumford-Shah model (1985) for image segmentation described by image approximations by a step (piecewise constant) function are described and developed to formalize...
The paper focuses on image segmentation by means of approaching by piecewise-constant approximations. The objective is the improvement of the segmented image, expressed in the noticeable drop of image approximation error (total squared error). The proposed method involves segment dividing into two in some image part and merge of a pair of adjacent segments in another image part, keeping a total segment number constant. For further enhancement of the segmentation, the proposed method is used in combination with advanced K-means method. The steep decline of the approximation error is provided along with obvious increase of perceptual quality of image segmentation. The effect is achieved owing to binary adaptive hierarchy of nested segments, generated for each segment. The segmentation improvement method is usable for the advancement of the computer vision systems using conventional segmentation by partitioning image into connected segments.
The paper refers to interdisciplinary research in the areas of hierarchical cluster analysis of big data and ordering of primary data to detect objects in a color or in a grayscale image. To perform this on a limited domain of multidimensional data, an NP-hard problem of calculation of close to optimal piecewise constant data approximations with the smallest possible standard deviations or total squared errors (approximation errors) is solved. The solution is achieved by revisiting, modernizing, and combining classical Ward’s clustering, split/merge, and K-means methods. The concepts of objects, images, and their elements (superpixels) are formalized as structures that are distinguishable from each other. The results of structuring and ordering the image data are presented to the user in two ways, as tabulated approximations of the image showing the available object hierarchies. For not only theoretical reasoning, but also for practical implementation, reversible calculations with pixel sets are performed easily, as with individual pixels in terms of Sleator–Tarjan Dynamic trees and cyclic graphs forming an Algebraic Multi-Layer Network (AMN). The detailing of the latter significantly distinguishes this paper from our prior works. The establishment of the invariance of detected objects with respect to changing the context of the image and its transformation into grayscale is also new.