MACHINE VISION AND IMAGE PROCESSING
Year:2016 | Volume:3 | Issue:1 |Papers Count:6

Identification and Verification using fingerprint has direct relation with quality of fingerprint. In this paper، novel method is proposed to enhance the quality of fingerprint using Arc-Gabor filter bank. Indeed proposed filter bank is an adaptive version of standard filter bank for fingerprint. Evaluating the ability of proposed method is done using two methods. In the first method، results of verification and identification using normalized histogram of Binarized Statistical Image Feature (BSIF) are compared for the fingerprints before and after using proposed method. And in the second method، Peak Signal to Noise Ratio (PSNR) is used. Two databases DBI and DBII are used for evaluation. Using proposed method، for verification Equal Error Rate، are decrease from (15. 89% and 11. 70%) to (11. 35% and 8. 00%)، respectively، and for identification، 1st rank correct recognition rate are increased from (69. 28% and 71. 16%) to (78. 80% and 81. 70%)، respectively. Average values of PSNR for enhanced fingerprint using proposed methods are more than other version of Gabor filter.

Graphology is a technical knowledge that presented a person، character or mood. In this research، by obtaining handwriting and personality questionnaire MMPI from different people، we formed a database. A decision tree was exploited as a machine learning tool for grouping extracted features. The proposed system can predict individual characteristic of the writer with scanning the handwriting sample. The samples included 140 subjects in two different groups. These groups include 70 clients of Semnan prison and 70 students of Semnan University of Applied Sciences. In order to obtain the handwriting، a neutral and smooth text was chosen and wanted the subjects to write in a non-line of A4 paper. A decision tree is used for a final recognition. To verify the recognition of the decision tree، the output of MMPI questionnaire was used that yielded promising result. Results of extracted features accuracy obtained by the proposed method showed %87 to %94 match the observations. %69 personality recognition accuracy (result of the decision tree) achieved compared with the results obtained personality MMPI questionnaire.

Blur is one of the common image distortions in which fraction of a pixel value، depending on the severity of blur، is added to one or moreneighboring pixels. Knowledge about the severity of blurrinessin a given image is very importantfor image enhancementpurposes. Sinceblur severity is not generally constant at all pixels، it is necessary to accurately estimate the blurriness value of pixels in the image (the so called blur map). In this paper، a methodis proposed for this purpose. The proposed method in this paper provides the blur map via estimating the blurriness of the pixels based on the difference between surrounding block of the pixel and its blurred version. Our investigations indicate that، in general، the blur image regions have lower frequency content than the un-blur (sharp) regions. The results of extensive experiments on a large database including motion and defocus blurred images confirmthe superiority of the proposed approachover the existing methods.

This paper attempts to recognize online Farsi sub-words using the Freeman chain codes and hidden Markov model. Chain codes reduce the number of data with using the direction of breaks and keeping the direction of pen movement. Hence it can be used as an effective way to recognize of online sub-words. After breaking the sub-word into component parts (main body and strokes)، each part separately coded using Freeman chain code. Since these codes are not sufficient to recognize of sub-words، we merged them with some other features extracted from horizontal and vertical trajectories. Finally، the set of features are sent hidden Markov model classifier to final recognition. Modeling has been built with Baum-Welch algorithm and training of program performed with forward algorithms. Using the mentioned steps on a database including the 2000 sub-words has the recognition rate of 93. 5%.

Persian specific Attribute for Handwritten Image Word spotting and recognition by Label-EmbeddingOCR methods cannot handle degraded printed documents and handwritten documents، Word spotting is an alternative way to search for query words. Nowadays، efficient Attribute-based Content Based Image Retrieval (CBIR) are developed for Image documents. In this paper new attributes، based on Persian writing Style are introduced for Persian word spotting. We compared and analyze our methods against the other state of the art methods. In addition، the proposed method can handle both Image and string queries، also it can be employed as a word recognition system. Experiments on two standard datasets، Farsa and Iranshahr، shows reasonable results.

Measurement of industrial parts such as gear is accomplished by various contact and non-contact methods. In this research dimensional inspection of the spur gear was developed through a non-contact method، namely “ Machine Vision” ، to find out the errors of general design parameters of gear as well as its teeth profile errors. For this purpose، a software was developed in LabVIEW environment. This software after acquiring the image of gear and implementing some low-level processes such as low-pass filtering، look up table transformation and image calibration، by performing some high-level process such as particle analysis، edge detection and geometrical operation can firstly finds out the gear location in the image and then can calculate it's geometrical parameters. This method has some advantages such as low cost، high speed، practicability of on-line measurement، ability to intelligent detecting the dimensional parameters of gear without need to importing data.