Page 208 - Kaleidoscope Academic Conference Proceedings 2024
P. 208

2024 ITU Kaleidoscope Academic Conference




















                                  Figure 2 – Semantic-aided image transfer through multi-modality


           semantic representation. Efficient transmission via channel
           encoding follows.
           Semantic Decoding:  Decode the transmitted semantic
           data at the receiver’s end using similar models as in the
           encoding phase to interpret the content and control signals,
           reconstructing an image representation.
           Image Interpretation: Interpret the reconstructed image
           representation and accompanying textual descriptions to
           visualize and understand the conveyed content, completing
           the semantic-aided image transfer through multi-modality.
           We assumed a noiseless channel in our discussion to facilitate
           exploration of data reduction techniques in semantic-aided
           image transfer through multi-modality.
                                                               Figure 3 – Average MSE for each mode of communication
                             3. RESULTS

           In our study, we conducted a comprehensive evaluation
           of various communication modes, considering three key
           performance metrics: MSE, PSNR and SSI. This evaluation
           encompasses both qualitative and quantitative aspects,
           providing a thorough analysis of the effectiveness and
           reliability of each communication mode.

           3.1 Why is Line-Art the best as a second mode?

           3.1.1 MSE

           The MSE averages for each communication mode, as depicted
           in Figure 3, reveal that the dual modality outperforms the
                                                              Figure 4 – Average PSNR for each mode of communication
           single modality. A smaller MSE value suggests superior
           reconstruction. Notably, within the dual modality, lineart  ability to maintain signal quality. A higher SSI indicates
           mode excels, emphasizing the importance of mode selection.  that the image is more perceivable by a human observer.
                                                              Unlike MSE and PSNR, which are quantitative approaches to
           3.1.2  PSNR                                        evaluate reconstructed images, SSI focuses on the qualitative
                                                              aspect of evaluation, reflecting the visual fidelity and
           The average PSNR for each communication mode, as shown  similarity to the original image.
           in Figure 4, highlights line art as the optimal choice for  As demonstrated by various results, line art consistently
           the second mode, emphasizing its superior performance in  exhibits the highest SSI, lowest MSE, and superior
           preserving signal quality. The reconstruction of the image is  PSNR compared to other modes. Hence, it unequivocally
           considered better when the value of PSNR is higher.  emerges as the optimal choice for the second mode of
                                                              communication, ensuring superior signal preservation and
           3.1.3  SSI                                         overall performance.
                                                              Table 1 presents a comprehensive analysis of various modes
           The average SSI for each communication mode, is illustrated  of communication, illustrating their respective results through
           in Figure 5. The graph underscores lineart as the prime  three informative examples.
           selection for the second mode, emphasizing its superior




                                                          – 164 –
   203   204   205   206   207   208   209   210   211   212   213