子計畫二：應用於智慧型行動載具之隱私保護與高效率壓縮研究

In this study, we present a privacy protection and high-efficiency compression technique for medical images in order to be applied to mobile intelligent carriers. First, we present an improved watermarking technique for the authorization of hospital logos. Experimental results show that the proposed technique has lower computational complexity and higher robustness than previous methods. Next, we present a new SPIHT based reversible watermarking and compression for medical images. By using a bit-stream insertion scheme, the patient’s information can be embedded into the bit-stream of a cover image for remote transmission. Experimental results show that the compression ratio can be up to 3.025. The watermarking capacity can be achieved over 0.75 bpp for some common images. In addition, the watermark can be extracted exactly and the cover image can be reconstructed either in a lossless way or a lossy way. The obtained results also show the improvement with respect to previous works. Finally, we take some situations in hospitals into account to present an improved Kinect midair handwritten number recognition technique. The recognition accuracy can be improved from 86.7% to 94.6%. By using this recognition technique, we create two Android apps to extend the utility of the technique.