catalog.txt

NIST Handwriting OCR Testbed

 blob_split_n_test - splits an image along a specified vertical path and then
              tests to see if any of the resulting pieces is empty or too small
              in which case cut points are adjusted and a new path is found.
 find_cut_path - given a starting x position, finds a vertical cut path through
              the image by following the gradient of character stroke edges.
 split_blob_along_path - given a vertical path, actually divides the image into
              two separate pieces.
 join_2_vert_paths - joins to vertical paths at their closest points and ensures
              that there is one and only one point in the resulting path for each
              y-value in the image.
 cut_caused_fragments - determines if cut created spurious fragments of characters
              by counting the number of non-dot blobs in the resulting pieces.
 blob_cut_x_as_pair - use multiple character detector to select an optimal
                      vertical cut point in an image of two touching characters.
 blob_cut_x_fr_left - use multiple character detector to select an optimal
                      vertical cut point working from left-to-right in an image
                      containing multiple touching characters.
 is_dot_blob - determines if blob is small enough to be a dot based on the
               estimated stroke width.
 process_dot_blobs - reconnects dots to i's and j's and other larger blobs that are
               sufficiently close.
 left_closer_to_dot - given two nbring blobs, determines which is closer.

 reconnect_dot - determines if nbring blob is below, close, and in line with the
                 dot blob.
 dist_dot2blob - computes the perpendicular distance from the blob to the linear
                 projection from a nbring blob.
 count_nondots_blob - counts the number of non-dots (blobs larger than a dot) in a
                 specified blob image.
 find_merge_blobs - determines if a blob should be merged with another nbring
                    blob based on horizontal overlap.
 detect_multi_char_blob - determines if the blob passed contains multiple
              characters based on a simple 2D-linear discrimination function.
 detect_multi_alpha_blob - determines if the blob passed contains multiple
    alphabetic characters based on a simple 2D-linear discrimination function.
 is_noise_blob - determines if blob is small enough to be noise based on the
                 estimated stoke width.
 process_noise_blobs - small blobs are merges back with larger blobs that are
                 sufficiently close.
 paste_top_strokes_blobls - loops through a blob list and puts top
                    horizontal strokes back on characters such as 5s.
 paste_alpha_tops - merges the tops of alphabetic characters with their
                    corresponing bodies.

 segment_chars8 - takes a binary charimage of a field and segments it into
                  characters by carefully compositing and spliting blobs.
 blob2chars8 - takes a list of blobs and the run lengths from a field and
               constructs single character images.
 compose_chars8 - takes a list of blobs and attempts to reconstruct characters that
               are fragmented, place dots back on i's and j's, and identify other
                  small blobs as noise.
 split_chars8 - takes a list of blobs and attempts to detect and split multiple
                touching characters.
 blobs2alphas8 - processes a list of blobs into a list of alphabetic
                 characters.
 compose_alphas8 - merge blobs of characters pieces into composite single
                   characters.
 split_alphas8 - takes a list of blobs and attempts to detect and split multiple
                touching alphabetic characters.
 segment_paragraph_blobls - processes a list of blobs extracted from a
             binary char image of a free-format handprinted paragraph
             into a list of character images.
 one_like_blob - determines if the given blob is shaped similar
                 to the handprinted character 1.
 one_like_blob - determines if the given blob is shaped similar
                 to a handprinted horizontal dash.
 is_smallpunct_blob - determines if the given blob is similar in
                 size and shape to a small punctuation mark.
 is_top_of_5 - determines if the given blob is shaped and positioned
               similar to the top stroke of a handprinted 5.
 is_top_of_alpha - tests if blob is shaped like a top stroke of alphabetic
                   character.
 split_touching_chars - separates multiple touching characters into individual
                        characters.
 split_touching_alphas - separates multiple touching characters into individual
                        alphabetic characters.
 segment_blob_as_pair - attempts to split touching character image based on
                        assumption that only two charcters are in the image.
 segment_blob_fr_left - attempts to split off a character from within the image
                        of touching charcters by working from left-to-right.