Image Retrieval based on Sectional Intensity Tolerance Measure for Color using Fuzzy Rule Sets

To solve the problem of false indexing and incorrect retrieval of color images, we propose a naval approach for color image retrieval which uses various features of different sections of the image. The proposed method uses different sizes of box filters to generate many numbers of subsamplingimages of differents ections and extracts the features of the image. The subsampled images are used to generate feature descriptors of those regions of interest. Generated descriptors represented the features of the image and used to produce a feature vector of the selected region. For each feature vector a rule is generated, and each rule has N number of intensity approximated value which depends on the number of feature descriptors generated for any specific region. For each section of the image made, a single tolerance value is generated. The generated features are used to produce a single value to represent the image and based on the fuzzy using rule set available the image class is identified. The proposed retrieval approach has produced efficient results.