public class L_Recog
extends com.sun.jna.Structure
This is a simple utility for training and recognizing individual
machine-printed text characters. It is designed to be adapted
to a particular set of character images; e.g., from a book.
There are two methods of training the recognizer. In the most
simple, a set of bitmaps has been labeled by some means, such
a generic OCR program. This is input either one template at a time
or as a pixa of templates, to a function that creates a recog.
If in a pixa, the text string label must be embedded in the
text field of each pix.
If labeled data is not available, we start with a bootstrap
recognizer (BSR) that has labeled data from a variety of sources.
These images are scaled, typically to a fixed height, and then
fed similarly scaled unlabeled images from the source (e.g., book),
and the BSR attempts to identify them. All images that have
a high enough correlation score with one of the templates in the
BSR are emitted in a pixa, which now holds unscaled and labeled
templates from the source. This is the generator for a book adapted
recognizer (BAR).
The pixa should always be thought of as the primary structure.
It is the generator for the recog, because a recog is built
from a pixa of unscaled images.
New image templates can be added to a recog as long as it is
in training mode. Once training is finished, to add templates
it is necessary to extract the generating pixa, add templates
to that pixa, and make a new recog. Similarly, we do not
join two recog; instead, we simply join their generating pixa,
and make a recog from that.
To remove outliers from a pixa of labeled pix, make a recog,
determine the outliers, and generate a new pixa with the
outliers removed. The outliers are determined by building
special templates for each character set that are scaled averages
of the individual templates. Then a correlation score is found
between each template and the averaged templates. There are
two implementations; outliers are determined as either:
(1) a template having a correlation score with its class average
that is below a threshold, or
(2) a template having a correlation score with its class average
that is smaller than the correlation score with the average
of another class.
Outliers are removed from the generating pixa. Scaled averaging
is only performed for determining outliers and for splitting
characters; it is never used in a trained recognizer for identifying
unlabeled samples.
Two methods using averaged templates are provided for splitting
touching characters:
(1) greedy matching
(2) document image decoding (DID)
The DID method is the default. It is about 5x faster and
possibly more accurate.
Once a BAR has been made, unlabeled sample images are identified
by finding the individual template in the BAR with highest
correlation. The input images and images in the BAR can be
represented in two ways:
(1) as scanned, binarized to 1 bpp
(2) as a width-normalized outline formed by thinning to a
skeleton and then dilating by a fixed amount.
The recog can be serialized to file and read back. The serialized
version holds the templates used for correlation (which may have
been modified by scaling and turning into lines from the unscaled
templates), plus, for arbitrary character sets, the UTF8
representation and the lookup table mapping from the character
representation to index.
Why do we not use averaged templates for recognition?
Letterforms can take on significantly different shapes (eg.,
the letters 'a' and 'g'), and it makes no sense to average these.
The previous version of this utility allowed multiple recognizers
to exist, but this is an unnecessary complication if recognition
is done on all samples instead of on averages.
| Modifier and Type | Class and Description |
|---|---|
static class |
L_Recog.ByReference |
static class |
L_Recog.ByValue |
| Modifier and Type | Field and Description |
|---|---|
int |
ave_done
set to 1 when averaged bitmaps are made
C type : l_int32 |
L_Bmf.ByReference |
bmf
bmf fonts
C type : L_Bmf* |
int |
bmf_size
font size of bmf; default is 6 pt
C type : l_int32 |
com.sun.jna.ptr.IntByReference |
centtab
table for finding centroids
C type : l_int32* |
int |
charset_size
expected number of classes in charset
C type : l_int32 |
int |
charset_type
one of L_ARABIC_NUMERALS, etc.
C type : l_int32 |
L_Rdid.ByReference |
did
temp data used for image decoding
C type : L_Rdid* |
L_Dna.ByReference |
dna_tochar
index-to-char lut for arbitrary charset
C type : L_Dna* |
int |
linew
use a value > 0 to convert the bitmap
C type : l_int32 |
float |
max_ht_ratio
max of max/min template height ratio
C type : l_float32 |
int |
max_splith
max component height kept in splitting
C type : l_int32 |
float |
max_wh_ratio
max width/height ratio to split
C type : l_float32 |
int |
maxarraysize
initialize container arrays to this
C type : l_int32 |
int |
maxheight_u
max height averaged unscaled templates
C type : l_int32 |
int |
maxwidth
max width averaged scaled templates
C type : l_int32 |
int |
maxwidth_u
max width averaged unscaled templates
C type : l_int32 |
int |
maxyshift
vertical jiggle on nominal centroid
C type : l_int32 |
int |
min_nopad
min number of samples without padding
C type : l_int32 |
int |
min_splitw
min component width kept in splitting
C type : l_int32 |
int |
minheight_u
min height averaged unscaled templates
C type : l_int32 |
int |
minwidth
min width averaged scaled templates
C type : l_int32 |
int |
minwidth_u
min width averaged unscaled templates
C type : l_int32 |
Numaa.ByReference |
naasum
area of all (scaled) templates
C type : Numaa* |
Numaa.ByReference |
naasum_u
area of all unscaled templates
C type : Numaa* |
Numa.ByReference |
nasum
area of (scaled) averaged templates
C type : Numa* |
Numa.ByReference |
nasum_u
area of unscaled averaged templates
C type : Numa* |
int |
num_samples
number of training samples
C type : l_int32 |
Pixa.ByReference |
pixa
averaged (scaled) templates per class
C type : Pixa* |
Pixa.ByReference |
pixa_id
input images for identifying
C type : Pixa* |
Pixa.ByReference |
pixa_tr
all input training images
C type : Pixa* |
Pixa.ByReference |
pixa_u
averaged unscaled templates per class
C type : Pixa* |
Pixaa.ByReference |
pixaa
all (scaled) templates for each class
C type : Pixaa* |
Pixaa.ByReference |
pixaa_u
all unscaled templates for each class
C type : Pixaa* |
Pixa.ByReference |
pixadb_ave
unscaled and scaled averaged bitmaps
C type : Pixa* |
Pixa.ByReference |
pixadb_boot
debug: bootstrap training results
C type : Pixa* |
Pixa.ByReference |
pixadb_split
debug: splitting results
C type : Pixa* |
Pix.ByReference |
pixdb_ave
debug: best match of input against ave.
C type : Pix* |
Pix.ByReference |
pixdb_range
debug: best matches within range
C type : Pix* |
Pta.ByReference |
pta
centroids of (scaled) ave.
|
Pta.ByReference |
pta_u
centroids of unscaled ave.
|
Ptaa.ByReference |
ptaa
centroids of all (scaledl) templates
C type : Ptaa* |
Ptaa.ByReference |
ptaa_u
centroids of all unscaled templates
C type : Ptaa* |
L_Rch.ByReference |
rch
temp data used for holding best char
C type : L_Rch* |
L_Rcha.ByReference |
rcha
temp data used for array of best chars
C type : L_Rcha* |
Sarray.ByReference |
sa_text
text array for arbitrary charset
C type : Sarray* |
int |
scaleh
scale all examples to this height;
C type : l_int32 |
int |
scalew
scale all examples to this width;
C type : l_int32 |
int |
setsize
size of character set
C type : l_int32 |
com.sun.jna.ptr.IntByReference |
sumtab
table for finding pixel sums
C type : l_int32* |
int |
templ_use
template use: use either the average
C type : l_int32 |
int |
threshold
for binarizing if depth > 1
C type : l_int32 |
int |
train_done
set to 1 when training is complete or
C type : l_int32 |
| Constructor and Description |
|---|
L_Recog() |
L_Recog(com.sun.jna.Pointer peer) |
| Modifier and Type | Method and Description |
|---|---|
protected java.util.List<?> |
getFieldOrder() |
allocateMemory, allocateMemory, autoAllocate, autoRead, autoRead, autoWrite, autoWrite, cacheTypeInfo, clear, ensureAllocated, equals, fieldOffset, getAutoRead, getAutoWrite, getFieldList, getFields, getNativeAlignment, getNativeSize, getNativeSize, getPointer, getStringEncoding, getStructAlignment, hashCode, newInstance, newInstance, read, readField, readField, setAlignType, setAutoRead, setAutoSynch, setAutoWrite, setFieldOrder, setStringEncoding, size, sortFields, toArray, toArray, toString, toString, useMemory, useMemory, write, writeField, writeField, writeFieldpublic int scalew
public int scaleh
public int linew
public int templ_use
public int maxarraysize
public int setsize
public int threshold
public int maxyshift
public int charset_type
public int charset_size
public int min_nopad
public int num_samples
public int minwidth_u
public int maxwidth_u
public int minheight_u
public int maxheight_u
public int minwidth
public int maxwidth
public int ave_done
public int train_done
public float max_wh_ratio
public float max_ht_ratio
public int min_splitw
public int max_splith
public Sarray.ByReference sa_text
public L_Dna.ByReference dna_tochar
public com.sun.jna.ptr.IntByReference centtab
public com.sun.jna.ptr.IntByReference sumtab
public Pixaa.ByReference pixaa_u
public Ptaa.ByReference ptaa_u
public Numaa.ByReference naasum_u
public Pixaa.ByReference pixaa
public Ptaa.ByReference ptaa
public Numaa.ByReference naasum
public Pixa.ByReference pixa_u
public Pta.ByReference pta_u
public Numa.ByReference nasum_u
public Pixa.ByReference pixa
public Pta.ByReference pta
public Numa.ByReference nasum
public Pixa.ByReference pixa_tr
public Pixa.ByReference pixadb_ave
public Pixa.ByReference pixa_id
public Pix.ByReference pixdb_ave
public Pix.ByReference pixdb_range
public Pixa.ByReference pixadb_boot
public Pixa.ByReference pixadb_split
public L_Bmf.ByReference bmf
public int bmf_size
public L_Rdid.ByReference did
public L_Rch.ByReference rch
public L_Rcha.ByReference rcha
Copyright © 2018. All Rights Reserved.