000			03170nam a22001337a 4500
999			_c66479 _d66476
100			_aFarah Mujtaba _a15MSCS12 _aSupervisor Dr. Akhter Hussain Jalbani
245			_aHand Face Blob Image Segmentation Using Trainable Weka Segmentation
260			_aNawabshah: _bQUEST, _c2019.
300			_a63p.
500			_aABSTRACT Artificial intelligence (Al) is the field marked ' intelligent agents ' himself. Al is part of computer science related to design computer systems that show human Intelligence. Computer vision is area of artificial intelligence which extract information from images by analyzing, mining, modeling, segmentation and understanding. Image segmentation is the procedure of partitioning of a digital image into multiple tracks (sets of pixels, also known as Super pixels) has been simplified for the purpose of segmentation and/or Represents an image that is more meaningful and easy to examine. Now a days image segmentation is a problem which is solved by many ways .Face segmentation of an image is a critical problem that acquire significance, Face recognition and segmentation play a major role in face recognition systems. There are many problems to be resolved so that make successful detection and the segmentation algorithm. This work creating segmentation scenario where Weka Trainable Segmentation (TWS) will be used in order to segment the images. Analyzing the performance of algorithms for segmentation based on feature selection of image, and training data set Segment the image into different areas and compare the results of different training features of trainable weka segmentation and provide the good result. Trainable Weka Segmentation (TWS) plugin of Fiji software which is used to segment the hand face blob image. There is many training feature in TWS for image segmentation which is compared and find good training feature for segmentation, Segment the image in 3 classes, background, hand and face. In this research make a seven set of different training feature and every set apply on each image and compare the result. According to this research, result find the accuracy of 15 images, SYG (Sobel, variance, Gaussian blur) training feature give 18.72 % result out of 15 images, Laplician , Minim um, 13ilteral (LMB) training feature give 34.20 % result out of 15 image, Difference of Gaussian blure, Maximum, Anistropic (OMA) training feature give 19.04% result out of 15 image, Hessian matrix, Median, Kuwahara (HMK) training feature give 5.98 % result out of 15 image, Gabor filter, Entropy, Billral (GEB) training feature give 21 .78 % result out of 15 image, Neighbour, Mean, Biletral (NMB) training feature give 71.94 % result out of 15 image, Neighbour, Mean, Bilteral, Structure (NM13S) training feature give 88.36 % result out of 15 image. NMBS Give good result out of 7 set of training features and HMK give worst result. Result summary of 15 images in percentage in which all over NM BS give 34 % highest percentage, NMB give 28%, GEl3 give 9 %, l l MK give 2 %, OMA give 7%,LMB give13%, SVG give 7 %.
700			_aDepartment of Information Technology
856			_uhttp://tinyurl.com/arwpa4dx
942			_cTHESIS