Study on Image Quality for Medical Ultrasonic Echo Image Compression by Wavelet Transform

Abstract

JPEG2000 is a new standard for image compression, which is based on wavelet transform and incorporated into DICOM (digital imaging and communication in medicine). In JPEG2000, there are two quantization methods, the scalar derived quantization (SDQ) and the scalar expounded quantization (SEQ). In this work, we use three optimization methods of quantization step, which are SDQ, SEQ determined by arithmetic average method and SEQ determined by genetic algorithm. Evaluation of image quality is comparison between peak signal to noise ratio (PSNR) and standard deviation, which one is correlation with the results of subjective assessment. PSNR is traditional objective assessment to measure image quality and standard deviation is new criterion. Subjective assessment is evaluated by ultrasonographers, which is the most important for medical image. The results show that standard deviation match to subjective assessment. Sometime compressed image loses important information for diagnosis in spite of high PSNR. Therefore, PSNR does not correlate well with the results of subjective assessment. Our purpose is to use the new criterion instead of PSNR that standard deviation is better than PSNR.