A Novel HMM Based Integrated Technique for Recognizing Clothes Patterns and Colors

Abstract

There is need to develop such system which will give automatically the pattern of cloth and the color of the cloth, which is selected for recognition. The image acquired by the camera and is further processed for recognizing color and pattern.  First we have to get the features of image that we selected for recognition and after that apply the Support vector Machine algorithm for further processing. In this implementation we used total three descriptors. To extract the statistical features we used Random signature descriptor in which there is use of DWT (Discrete Wavelet Transform) for calculating the global features of the cloths we selected. The obtained global features are then combined with local features which are calculated by the scale invariant feature transform to recognize the complex pattern of cloths. After getting the features we used SVM (support vector machine) to classify the images or pattern. In this we used CCNY dataset for recognizing cloth pattern. With the help of this method the physically impaired person can easily recognize the pattern of cloths as well as color of cloth. The MATLAB execution results show the performance of the system developed. Random Signature is used to capture the global directionality features statistical descriptor (STA) to extract the global statistical features on wavelet sub-bands and Scale Invariant Feature Transform (SIFT) to represent the local structural features. The combination of multiple feature channels provides complementary information to improve recognition accuracy. The   collection of dataset on clothing pattern recognition including four-pattern categories of plaid, striped, pattern less, and irregular, the method also provides new functions to improve the life quality for blind and visually impaired people. The obtained results are improved to great extent with the help of HMM (Hidden Markov Models).