Image Based Search System Using Hierarchical Object Category Recognition Technique

In this paper, we present the information search system using object category recognition, which is queried by image from mobile phone camera or from photo sharing service on internet. In such system, processing speed is an important requirement. We adopted “Standard Model” proposed by T. Serre in 2005, and improved processing speed by replacing Gabor filter to Haar wavelet, vector quantization of feature patch, and restriction of calculation area. In addition, by retaining the information of each feature’s position, it compensates the accuracy which is a little reduced in exchange of processing speed. We implemented this method to server system, and proved this system can work in practical processing time. Through the experiment for Caltech-101 image database, we confirmed value of this system.