Adaptive Region Growing Color Segmentation for Text Using Irregular Pyramid

This paper presents the result of an adaptive region growing segmentation technique for color document images using an irregular pyramid structure. The emphasis is in the segmentation of textual components for subsequence extraction in document analysis. The segmentation is done in the RGB color space. A simple color distance measurement and a category of color thresholds are derived. The proposed method utilizes a hybrid approach where color feature based clustering followed by detailed region based segmentation is performed. Clustering is done by merging image color points surrounding a color seed selected dynamically. The clustered regions are then put through a detailed segmentation process where an irregular pyramid structure is utilized. Dynamic and repeating selection of the most suitable seed region, fitting changing local condition during the segmentation, is implemented. The growing of regions is done through the use of multiple seeds growing concurrently. The algorithm is evaluated according to 2 factors and compared with an existing method. The result is encouraging and demonstrates the ability and efficiency of our algorithm in achieving the segmentation task.