Optimized End-to-End Coiflets Discrete Wavelet Transform for Dermoscopic Images Compression

Abstract

The dramatical increase in the number of images within hospitals and for consultation emerge the need for efficient medical image compression to store and transmit them. For preserving data integrity, lossless compression is used with medical images. Accordingly, the proposed system preserves the image quality while compressing the dermoscopic images using an end-to-end optimized Coiflet discrete wavelet transform (E-EOCDWT). The particle swarm optimization (PSO) was applied to determine the optimal values of the coefficients’ threshold and the level of decomposition of Coiflet discrete wavelet transform (CDWT) with minimizing the mean square error (MSE). The proposed compression system includes three main blocks, namely CDWT and inverse CDWT; entropy encoding and decoding by Huffman coding; and PSO. The CDWT decomposes the dermoscopic images into sub-band coefficients, then these coefficients are compared to the optimal threshold value (OTS). The coefficients having values less than OTS are set to zero, and then encoded using Huffman coding, while the coefficients above OTS are quantized. The Huffman decoding is used to decode the compressed bitstream. Finally, the dermoscopic images are reconstructed by applying the inverse CDWT. The proposed E-EOCDWT method was evaluated by calculating the structure similarity index measurement (SSIM), MSE, peak signal to noise ratio (PSNR), computational time (Ct), and compression ratio (CR). It achieved 41.6 dB, 1.4, 0.86, 29 sec, and 59% of PSNR, MSE, SSIM, Ct, and CR, respectively.