The uprising number of applications that involve very large images with resolutions greater than 30 000×30 000 raises major memory management issues. Firstly, the amount of data usually prevents such images from being processed globally and therefore, designing a global image partition raises several issues. Secondly, a multi-resolution approach is necessary since an analysis only based on the highest resolution may miss global features revealed at lower resolutions. This paper introduces the tiled top-down pyramidal framework which addresses these two main constraints. Our model provides a full representation of multi-resolution images with both geometrical and topological relationships. The advantage of a top-down construction scheme is twofold: the focus of attention only refines regions of interest which results in a reduction of the amount of required memory and in a refinement process that may take into account hierarchical features from previous segmentations. Moreover, the top-down model is combined with a decomposition in tiles to provide an accurate memory bounding while allowing global analysis of large images.

Applicative fields based on the analysis of large images must deal with two important problems. First, the size in memory of such images usually forbids a global image analysis hereby inducing numerous problems for the design of a global image partition. Second, due to the high resolution of such images, global features only appear at low resolutions and a single resolution analysis may loose important information. The tiled top-down pyramidal model has been designed to solve this two major challenges. This model provides a hierarchical encoding of the image at single or multiple resolutions using a top-down construction scheme. Moreover, the use of tiles bounds the amount of memory required by the model while allowing global image analysis. The main limitation of this model is the splitting step used to build one additional partition from the above level. Indeed, this step requires to temporary refine the split region up to the pixel level which entails high memory requirements and processing time. In this paper, we propose a new splitting step within the tiled top-down pyramidal framework which overcomes the previously mentioned limitations.

To cite this version:Romain Goffe, Luc Brun, Guillaume Damiand. Tiled top-down pyramids and segmentation of large histological images. Springer. In 8th IAPR -TC-15 Workshop on Graph-based Representations in Pattern Recognition (GBR'11), May 2011, Munster, Germany. 6658, pp.255-264, 2011 Abstract Recent microscopic imaging systems such as whole slide scanners provide very large (up to 18GB) high resolution images. Such amounts of memory raise major issues that prevent usual image representation models from being used. Moreover, using such high resolution images, global image features, such as tissues, do not clearly appear at full resolution. Such images contain thus different hierarchical information at different resolutions. This paper presents the model of tiled top-down pyramids which provides a framework to handle such images. This model encodes a hierarchy of partitions of large images defined at different resolutions. We also propose a generic construction scheme of such pyramids whose validity is evaluated on an histological image application.