Contour Image Indexing using Dyadic Wavelet Transform of One Dimensional Signal with Branch Compensation

Various methods exist for the indexing of images. In this work we represent an image by a 1D signal. From this 1D signal we extract the features using wavelet transform to index the images. The 1D representation of the contours would help to decrease the computational complexity of processing the 2D signal, thereby increase the speed of the process. In order to represent a 2D image in terms of a 1D signal, the contour of the image is used. Various contour following algorithms exist for tracing contours. Some contours may contain two or more paths. To handle such situations a new algorithm is proposed that takes into account all the branches available and represent them as a 1D signal. Also in the process of tracing a new method of dynamic mask is introduced by which the number of coefficients to represent a contour is further reduced by a large factor.