Implementation of wavelet based decomposition and reconstruction of an image using TMS320C6701

The discrete wavelet transform provides sufficient information both analysis and synthesis of the original image with a significant reduction in the computation time. There are two approaches for working on the above algorithm, one being by using two dimensional filters and the other one by using separable transforms that can be implemented using a one-dimensional filter on the rows first and then on the columns. In this research, we have implemented wavelet decomposition and reconstruction using a one-dimensional transform applied on the rows first and then the columns. For an N x M image size, we filter each row and then column with the analysis pair of low-pass and high-pass filters and down sample successively to obtain four bands after decomposition. Later, during image reconstruction, we up sample and filter each column and then row with a pair of synthesis low pass and high pass filters to obtain the original image size. This algorithm has been implemented in real-time by using the floating-point processor TMS320C6701 chip manufactured by Texas Instruments (TI) that is widely used for image processing applications. The wavelet transform has become the most powerful tool for still image analysis. Yet there are many parameters within a wavelet analysis and synthesis that govern the quality of the image. In this paper, we discuss the wavelet decomposition and reconstruction strategies for a two-dimensional signal and their implications on the reconstruction of the image. A pool of grey scale images has been wavelet transformed using a set of bi-orthogonal filters (wavelet filter bank) that undergoes the decomposition and reconstruction process.