Análise comparativa de técnicas de compressão aplicadas a imagens médicas usando ultrassom

Abstract

The area of image compression is of great importance for medical applications. Compression techniques have been studied for allowing to efficiently representing the data, reducing the required storage space and minimizing the demand on transmission through communication channels. The JPEG compression technique uses a two-dimensional discrete cosine transform and techniques Joint Photographic Experts Group 2000 (JPEG2000), Set Partitioning in Hierarchical Trees (SPIHT) and Embedded Zerotree Wavelet (EZW) use the two-dimensional wavelet transform. In this work, a comparative analysis is made using the metrics Mean Squared Error (MSE), Peak Signal to Noise Ratio (PSNR), Cross Correlation (CC) and Structural Similarity (SSIM) as well as an indirect assessment of computational effort through time processing. The analysis using the JPEG compression technique using blocks of different sizes allows us to conclude on the influence of the center frequency of the ultrasonic transducers in the details of the image so as to conclude that the use of blocks of size 8x8 and 16x16 pixels has better performance than the use of note less size. Additionally, in evaluating alternative quantization matrices proposed by Hamamoto, Veraswamy and Abu for JPEG compression technique is concluded that these introduce improvements in the quality of reconstructed ultrasound image when compared to traditional matrix quantization considered that given compression rate in comparison. Techniques for using the wavelet transform have been possible to observe a significant impact on the quality of the ultrasound image reconstructed on the basis of the filters used. Finally, it was concluded that the JPEG2000 algorithm has the best performance with respect to image quality and processing time compared with JPEG, EZW and SPIHT algorithms.