Color Encoding of Multiple Layout Levels.

An example image to demonstrate color encoding of multiple layout levels. The images show word level and text-line level segmentation.

Different Performance Measures

Example image to illustrate different performance measures. The left image shows two color coded document images. A pixel correspondence graph obtained from these images is shown on the right side. The nodes corresponding to the ground-truth segments are labeled 1-7, whereas the nodes in the segmented image are labeled a-i. Only significant edges are shown in the pixel correspondence graph. Based on the definitions given in Section II-C, the values of each performance measure for this example are given on the right side of the graph.

Literature Survey

Performance Evaluation and Benchmarking

of Six Page Segmentation Algorithms

•  This paper give an idea about performance of the famous six segmentation algorithms namely Dummy algorithm, X-Y Cut, Smearing, Whitespace analysis, Constrained text-line detection and Voronoi-diagram based algorithm in various types of input images.   
• This paper deliver a good method for evaluating The performance of Segmentation algorithm.  and  
• This paper concludes such a way that which algorithm is best for  which types of input images. For eg. the x-y cut and the smearing algorithms fail to segment a page in the presence of noise. The whitespace analysis algorithm is sensitive to the stopping rule and results in either over-segmentations of under-segmentations. The docstrum and the Voronoi algorithms tend to over-segment title and section headings if the font size is much different from body text in that page. The constrained text-line finding algorithm misses single-digit page numbers as it requires at least two components to make a line.

Document Image Analysis

The objective of document image analysis is to recognize the text and graphics components in images of documents, and to extract the intended information as a human would.