The main aim of this paper  is to provide a better reconstruction technique for image by solving the drawbacks that currently exist in the literary works. Thus, we intended to propose an efficient image reconstruction technique by exploiting the Haar DWT to remove cracks to input image and accurate threshold values to very high PSNR compared to input cracked image by improved particle swarm optimization used for better theshold values. Normally in image reconstruction, the input image which is corrupted by noise is given to the reconstruction process and obtained the reconstructed image with high quality. we exploit a renowned filtering technique to yield high quality reconstructed image. The low quality blurred image will be processed by our renowned filtering technique and obtain a reconstructed image with high contrast. Thus, we finally obtain a reconstructed image from our proposed technique with enhanced quality. To prove the performance of our proposed technique, the reconstructed image quality will be compared with the conventional techniques. Overall, our proposed technique can reconstruct the image more effectively by two phases (i) training phase  and (ii) investigation phase achieving higher image quality ratio than the conventional reconstruction techniques.