Non-subsampled Complex Wavelet Transform Based Medical Image Fusion

The paper presents a feature based medical image fusion approach for CT and MRI images. The directional features are extracted from co-registered CT and MRI slices using Non-Subsampled Dual Tree Complex Wavelet Transform (NS DT-CxWT). These features are combined using average and maxima fusion rules to create composite spectral plane. The new visually enriched image is reconstructed from this composite spectral plane by applying inverse transformation. Such fused images are evaluated for its visual quality using subjective and objective performance metrics. The quality of fused image is rated by three radiologists in subjective evaluation whereas edge and similarity based fusion parameters are computed to estimate the quality of fused image objectively. The proposed algorithm is compared with the state of the art wavelet transforms. It provides visually enriched fused images retaining soft tissue texture of MRI along with bone and lesion outline from CT with better contrast for lesion visualization and treatment planning. It is also found that the average score by radiologists is ‘3.85’ for proposed algorithm which is much higher than that of the average score for other wavelet algorithms.