Multispectral code excited linear prediction coding and its application in magnetic resonance images

Area of Science:

• Medical Imaging
• Image Compression
• Signal Processing

Background:

• Multispectral images require efficient compression techniques.
• Existing methods like JPEG and EZW have limitations in preserving multispectral data quality.
• Previous work includes the multispectral segmented autoregressive moving average (MSARMA) method.

Purpose of the Study:

• To develop and evaluate a novel multispectral image compression method.
• To compare different linear prediction models and adaptation schemes.
• To optimize compression for multispectral magnetic resonance (MR) images.

Main Methods:

• A forward adaptive autoregressive (AR) model was employed for multispectral code excited linear prediction (MCELP).
• The proposed method, MFCELP, utilizes 3-D macroblocks and micro-blocks with analysis-by-synthesis for excitation signal determination.
• Vector quantization was used to specify the error for high-quality medical image reconstruction.

Main Results:

• The MFCELP method demonstrated a good balance of performance, complexity, and robustness.
• Significant visual improvements were observed compared to JPEG, EZW, VT, and MSARMA methods.
• The method was successfully applied to a large dataset of multispectral MR neuro images.

Conclusions:

• MFCELP is a highly effective method for compressing multispectral images, particularly MR images.
• The approach offers superior visual quality and compression efficiency over existing techniques.
• This method holds promise for enhancing medical image storage and transmission.