Text, photo, and line extraction in scanned documents

Combined method for scanned documents images segmentation using sequential extraction of regions

“…Methods for the segmentation of images of scanned documents are categorized in the scientific literature into block processing methods, pixel processing methods, and methods for processing connected components [3][4][5][6][7][8][9][10][11][12].…”

“…Authors in [9], in order to segment the images of scanned documents into text, photographs and separators in the form of lines, sequentially applied the pixel and block processing using 5 modules. The first module involved the preprocessing of images of scanned documents, which included scaling, image improvement, and a transition from the RGB color space to the CIE La*b* space.…”

“…In the fifth module, the results of extracting the text, illustrations, and separators in the form of individual maps for each class of segments, were combined using the method of k-means, thereby forming the result of segmenting the original image. Method [9] is distinguished by the high quality of segmentation, but low speed.…”

“…Known methods for the segmentation of images of scanned documents [2][3][4][5][6][7][8][9][10][11][12] are not capable to simultaneously satisfy modern requirements to the quality of segmentation and to the high-speed performance of segmentation of images of scanned documents. Block processing methods [3,4] are characterized by high-speed performance, but the result of segmentation depends on the partition of the processed image into blocks.…”

“…Block processing methods [3,4] are characterized by high-speed performance, but the result of segmentation depends on the partition of the processed image into blocks. At the same time, pixel processing [5][6][7][8][9] requires a significant time cost although it makes it possible to obtain high quality of segmentation. For methods that involve analysis of connected components [2,[10][11][12], the processing time and the quality of segmentation are determined based on the result of the identification of connected components.…”