Discrete Wavelet Transforms for PET Image Reduction/Expansion (wavREPro)

Abstract : The large volume of medical images remains a major problem for their archiving and transmission. In this context, we propose a novel protocol wavREPro that aims to minimize the image size before its storage and then to enlarge it at reception. This process has been achieved by exploiting the Discrete Wavelet Transforms (DWT) namely Haar, Le Gall (5/3) and Daubechies (9/7) wavelets. These tools represent the image in the multi-resolution domain that follows the human psycho-visual system. Therefore, the reduced image is none other than the approximation of the image. Its magnification is carried out by either cancelling the details (wavREProZ) or estimating them (wavREProED) using the DWT−1 on the reduced image. Our experiments have been conducted on a PET (Positron Emission Tomography) medical image database and the results have been presented for the three well-known color spaces RGB, HSV and YCbCr. The reported results have promoted the wavREProZ application with the Haar wavelets on RGB images since it achieved maximum fidelity between the original and reduced then enlarged images. The good performance of this approach encourages its adoption to display images on screens having different sizes.
Document type :
Book sections
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-01648634
Contributor : René Chalon <>
Submitted on : Sunday, November 26, 2017 - 8:34:21 PM
Last modification on : Thursday, November 21, 2019 - 2:22:33 AM

Identifiers

Citation

Mohamed Salim Bouhlel, Ali Khalfallah, Hedi Amri, Malek Gargouri, Mohamed Karim Abdmouleh, et al.. Discrete Wavelet Transforms for PET Image Reduction/Expansion (wavREPro). Balas V., Jain L., Balas M. Soft Computing Applications - Proceedings of the 7th International Workshop Soft Computing Applications (SOFA 2016), 633, Springer, pp.524-539, 2017, Advances in Intelligent Systems and Computing, 978-3-319-62521-8. ⟨10.1007/978-3-319-62521-8_45⟩. ⟨hal-01648634⟩

Share

Metrics

Record views

298