• G.R. Gunashree
• M. Kempanna

Transition region-based approaches are hybrid segmentation techniques well known for its simplicity and effectiveness. The segmentation effectiveness depends on robust extraction of transition regions. This paper proposes a transition region extraction method for image segmentation. The image is converted into wavelet domain by applying discrete wavelet transform. Standard deviation filtering and thresholding operations are applied to the image for extracting the transition region feature matrix. This feature matrix is used to find corresponding prominent wavelet coefficients. Inverse wavelet transform is applied to the coefficients to get the edge image with more than one-pixel width.Transition regions are got when global thresholding is applied. Morphological thinning and region filling are used to dilate thick edges and fill the holes. Using object regions,the objects are extracted and thus segmented. Using the trained convolution network these segmented images are classified. Then when we also perform a performance analysis for the segmentation performed. This is helpful for farmers in remote areas to detect the type of bug and take necessary actions and get better yield of crops

Transition region-based approach, Segmentation, CNN, performance.

"Segmentation and Classification: Separation ofWorm Image from Leaf", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 5, page no.324-327, May-2019, Available :http://www.jetir.org/papers/JETIRCD06057.pdf

"Segmentation and Classification: Separation ofWorm Image from Leaf", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 5, page no. pp324-327, May-2019, Available at : http://www.jetir.org/papers/JETIRCD06057.pdf