Font Size: a A A

Study On Image Retrieval Of Scanned Hand-painted Plant Image

Posted on:2014-09-20Degree:MasterType:Thesis
Country:ChinaCandidate:J YuFull Text:PDF
GTID:2268330422952533Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Plant is very important in the nature, and is inextricably linked with the productionof human life. The plant can maintain the composition of the atmosphere, promote thedevelopment and evolution of the Earth and the biosphere, maintain the ecologicalbalance, and provide energy for life on Earth. Plant taxonomy is of great importanceto research on rational use of the plant, and therefore carrying out the study of plantimages has a wide range of practical significance.This paper conceives ways of retrieval from the images of real colorful plant leavesto the images of scanned hand-painted plant leaves. And the author discusses indetails how to get the shapes and textures of the two types of plant images accordingto their different features. With experiments, set up a retrieval prototype by choosingpractical and feasible characteristic elements.The article mainly focuses on the following things:(1)establishes a sampledate-base of images of scanned hand-painted plant leaves (2)analyzes the methods ofpreprocessing the images of real colorful plant leaves and the images of scannedhand-painted plant leaves, including the normalize, filtering, binarizing andmorphological processing.(3)experiments commonly used ways of extracting shapesof leaves and bring forward an improved PHOG ways of describing blade shapes.(4)analyzes ways of describing textures and put forward a way of describing textureof image of scanned hand-painted plant leaves by using edge gray histogram andimproved PHOG.(5)studies the image retrieval technologies both based on shapesand texture features, and realized the weighted retrieval by both shape and texturefeatures. And last but not the least, the article explores the retrieval possibilities of thesketches of hand-painted leaves and the images of scanned hand-painted leaves.This article uses two databases to validate the algorithm. The first is the previouslymentioned self-built image database of scanned self-painted leaves, functioning as thesample database for the retrieval system and the second database is the ICL leavesfrom the Chinese Academy of Sciences Intelligent Computing Laboratory database in Hefei, functioning as an database for sample images.The retrieval program mentioned in the article provides a strong attempt for thestudy of plant classification and retrieval. With the ways employed by "Flora" to getbook illustrations of plant leaves, we can set up a close to a complete plant leaf imagelibrary and help to make retrieving the images of plant leaves more practical.
Keywords/Search Tags:images of scanned hand-painted leaves, Shape feature, Texture-feature, retrieval system, PHOG
PDF Full Text Request
Related items