Image compression and reconstruction by examplar based inpainting using wavelet transform on textural regions

Image compression is the method to decrease the redundant information of the image to store and transmit in a cost effective way. A novel image compression technique is proposed using discrete wavelet transformation (DWT) along with examplar based image inpainting. This technique is aimed to conserve the textural features of the image. Image is classified into featured region and non-featured region initially and also the regions that are texturally important are captured for assigning textural descriptor as standard and fuzzy clustering methods. The descriptors of textural region incorporate coexistence matrices deployed measures along with features obtained by coherence analysis. Discrete wavelet transform is applied to these regions distinctly where partitioning of the input image is done and estimation of number of wavelet coefficients is decided depending on the nature of textural clustering. Here different compression ratios are achieved for separate regions. Every reconstructed regions are linearly combined together to reconstruct the original image. Experiments are conducted on various images to test the proposed approach efficiency. It is found that the proposed method achieves larger coding performance and reconstructs the compressed image similar to the original image in a visually plausible way.