Blob analysis functions are used to segment the image and extract the characters constituting the text to be read. Blobs are elected as characters based on tunable size and shape criteria. Moreover, EasyOCR is able to deal with characters which are split into several blobs. When the exact position of the characters in the image is unknown, EasyOCR functions will process the entire image and locate the characters. 

Recognition process

EasyOCR follows a few steps in the recognition process.

First, the image is segmented, i.e. threshold and decomposed into objects or blobs (connected components), in the same way as EasyObject does.

Then the objects are filtered according to the size and possibly grouped together ("repasted") to form distinct characters. This is called character isolation or segmentation. When several characters touch each other, they can also be separated. This is called character cutting. The segmentation step can be bypassed when the exact position of the characters is known beforehand.

The characters are compared to a set of patterns, called a font. A character is recognized by finding the best match between a character and the patterns in the font.

Raw image

After segmentation

After character isolation

After recognition

The recognition process consists of the following elements:

• Read a pre-recorded font from a disk file;
• Segment the image to locate the characters;
• Select the objects considered as characters and sorts them from left to right;
• Perform the matching from object to characters

Recognition parameters

The recognition process is governed by a few parameters that need to be fine tuned to obtain the most reliable results.

• The following two parameters are used during segmentation:

- TextColor : black text on a white background, or conversely, with or without thresholding;

- Threshold value used to separate the text from the background. The threshold value should be chosen such that the characters are well separated.

• The following geometric parameters are used during character isolation:

- RemoveBorder: most of the time, blobs that are found along the image/ROI edges are spurious and cannot be exploited for character recognition. By default they are discarded for character isolation;

- NoiseArea: if a blob has an area smaller than this value, it is considered as noise and discarded. The NoiseArea should be chosen such that the noise blobs are discarded but small character features are preserved (f.i., the dot over an "i" letter);

- MaxWidth, MaxHeight: if a blob does not fit within a rectangle with these dimensions, it is not considered as a possible character (too large) and is discarded. Furthermore, if several blobs fit in a rectangle with these dimensions, they are grouped together, forming a single character. The outer rectangle size should be chosen such that it can contain the largest character from the font, enlarged by a small safety margin;

- MinWidth, MinHeight: if a blob or a group of blobs does fit in a rectangle with these dimensions, it is not considered as a possible character (too small) and is discarded. The inner rectangle size should be chosen such that it is contained in the smallest character from the font, shrunk by a small safety margin;

- RemoveNarrowOrFlat: by default, small characters are discarded when they both narrow and flat. This behavior can be changed so that they are discarded when either condition is met.

- Spacing: if to blobs are separated by a vertical gap wider than this value, they are considered to belong to different characters. This feature is useful to avoid the grouping of thin characters that would fit in the outer rectangle. Its value should be set to the width of the smallest gap between adjacent letters. If it is set to a large value (larger than MaxWidth), it has no effect.

- CutLargeChars: when a blob or grouping of blobs is larger than the maximum allowed width, it is considered as clutter and discarded. When the CutLargeChars mode is enabled, the blob is split in as many parts as necessary to fit. This is an attempt to separate touching characters;

- RelativeSpacing: when the CutLargeChars mode is enabled, setting this value allows specifying the amount of white space that should be inserted between the split parts of the blobs;

Learning

EasyOCR is a multi-font character recognition library. This means that EasyOCR functions are able to recognize text printed using any character font, once it has been taught. Practically, during the learning process, characters are presented one by one to the system which analyzes them and builds a database called a font.

Only a few data are stored for each new character, they represent distinctive features of the character’s shape. This small database may be saved to disk and restored when needed.

During the learning process, each pattern gets an associated numerical value call its code (usually its ASCII code). A pattern also belongs to a character class.