Design And Implementation Of Gabor Filter Based Offline YorÙbÁ Handwritten Recognition System.

CHAPTER THREE
RESEARCH METHODOLOGY
3.1    DATA ACQUISTION
The Yoruba handwriting images used in this project are those were created for the purpose of this project. This database was only recently assembled by the author of this project, and before this there was no standard database for this field. The database consists of a collection of Yoruba characters images, each containing one character. The images come from Ten (10) different writers, mostly students. All the figures in this thesis showing Yoruba charactes are examples from this database.

Figure 3.2: Yoruba Handwriting image for letter P character
3.2    PREPROCESSING
The raw data is subjected to a number of preliminary processing steps to make it usable in the descriptive stages of character analysis. Pre-processing aims to produce data that are easy for the Yoruba handwritten character recognition systems to operate accurately. In this project, the Gabor Filter Algorithm was used to enhance the quality of the preprocessing image for better recognition as shown below:
The main objectives of preprocessing are:
    Noise reduction
    Normalization of the data
    Compression in the amount of information to be retained
In order to achieve the above objectives the following techniques are utilized in the preprocessing stage;
3.2.1    NOISE REDUCTION
The noise, which is introduced by the optical scanning devices or the writing instrument, causes disconnected line segments bumps and gaps in lines filled loops etc. The distortion which includes local variations rounding of corners dilation and erosion also a problem Prior to the character recognition, it is necessary to eliminate these imperfections. In this project, the filtering technique was used to reduce the noise.
The basic idea is to convolve a predefined mask with the image to assign a value to a pixel as a function of the gray values of its neighboring pixels. The linear spatial mask was used, where the intensities x(i,j) of the input image is transformed to the output image by;

Where a_kl is the weight of the gray levels of pixels of the mask at location (k, l). Filters can be designed for smoothing, sharpening, theresholding and contrast adjustment purposes by (Lee, 1987).