Implementation of Dictionary Learning-based Image Compression using Local Variance

DOI: https://doie.org/10.1213/Jbse.2024442109

Sumangala Gangadharamath Shekhariah, Sreedharamurthy Sameraya Kanthiah

Keywords:

Image Compression, Dictionary Learning, Sparse representation

Abstract:

 Compared to traditional transformed-based strategies, compression of pictures utilizing learned dictionaries employing the K-SVD algorithm has demonstrated greater results. However, applying K-SVD to image patches without taking into account the local characteristics of the image can cause rate-distortion performance to suffer. Image features vary within an image and such variations can be effectively used in adaptive image processing techniques. An important image characteristic such as the local variance of image intensity is one such feature, which can be utilized in adaptive image coding technique. In this work, we propose dictionary learning-based adaptive coding  based on the local variance of the image region to improve the R-D performance of the compression scheme. The technique is applied to typical test images. PSNR and SSIM are used to assess the performance of the suggested approach. According to the simulation results, the suggested technique shows improved performance at specified bit rates.

PDF